SPR-606: Calibration and Implementation of the AASHTO ... · PDF filethe AASHTO...
Transcript of SPR-606: Calibration and Implementation of the AASHTO ... · PDF filethe AASHTO...
e
Calibration and Implementation of the AASHTO Mechanistic-Empirical Pavement Design Guide in Arizona
SEPTEMBER 2014
Arizona Department of Transportation Research Center
SPR 606
Calibration and Implementation of
the AASHTO Mechanistic‐Empirical
Pavement Design Guide in Arizona SPR‐606
September 2014
Prepared by:
Michael I. Darter
Leslie Titus‐Glover
Harold Von Quintus
Biplab B. Bhattacharya
Jagannath Mallela
Applied Research Associates, Inc.
100 Trade Centre Dr., Suite 200
Champaign, IL 61820
Prepared for:
Arizona Department of Transportation
206 S. 17th Ave.
Phoenix, AZ 85007
in cooperation with
U.S. Department of Transportation
Federal Highway Administration
This report was funded in part through grants from the Federal Highway Administration, U.S. Department of
Transportation. The contents of this report reflect the views of the authors, who are responsible for the facts
and the accuracy of the data, and for the use or adaptation of previously published material, presented herein.
The contents do not necessarily reflect the official views or policies of the Arizona Department of Transportation
or the Federal Highway Administration, U.S. Department of Transportation. This report does not constitute a
standard, specification, or regulation. Trade or manufacturers’ names that may appear herein are cited only
because they are considered essential to the objectives of the report. The U.S. government and the State of
Arizona do not endorse products or manufacturers.
Technical Report Documentation Page
1. Report No. FHWA‐AZ‐14‐606
2. Government Accession No.
3. Recipient's Catalog No.
4. Title and Subtitle Calibration and Implementation of the AASHTO Mechanistic‐Empirical Pavement Design Guide in Arizona
5. Report Date September 2014 6. Performing Organization Code
7. Author Michael I. Darter, Leslie Titus‐Glover, Harold Von Quintus, Biplab B. Bhattacharya, and Jagannath Mallela
8. Performing Organization Report No.
9. Performing Organization Name and Address
Applied Research Associates, Inc. 100 Trade Centre Drive, Suite 200 Champaign, IL 61820
10. Work Unit No.
11. Contract or Grant No. SPR 000‐1(169) 606
12. Sponsoring Agency Name and Address Arizona Department of Transportation 206 S. 17th Ave. Phoenix, AZ 85007
13.Type of Report & Period CoveredFINAL REPORT 2008‐2012 14. Sponsoring Agency Code
15. Supplementary Notes Prepared in cooperation with the Federal Highway Administration. 16. Abstract This report documents efforts of the Arizona Department of Transportation (ADOT) to implement the American Association of State Highway and Transportation Officials (AASHTO) DARWin‐ME pavement design guide in Arizona. The research team also prepared a practical stand‐alone user’s guide that provides guidance for obtaining inputs, conducting design, and establishing the recommended pavement design. Implementation focused on identifying the desired pavement design application of flexible hot‐mix asphalt (HMA) pavements, composite pavements (thin asphalt rubber friction course over jointed plain concrete pavement [JPCP] and continuously reinforced concrete pavement [CRCP]), JPCP, and HMA overlays of flexible pavement; characterizing materials and subgrades; determining traffic loadings (conducted under Darter et al. 2010); collecting and assembling DARWin‐ME input data from 180 Long Term Pavement Performance and pavement management system sections of flexible, rigid, composite, and rehabilitated pavements; calibrating the DARWin‐ME distress and International Roughness Index (IRI) prediction models to Arizona conditions; and training ADOT staff. Several biased distress and IRI models were corrected through the local calibration of Arizona pavements. Several key inputs were more accurately defined and Arizona defaults provided (e.g., subgrade resilient modulus). The calibration process improved these models through verification, validation, and calibration with Arizona data. Overall, the inputs and calibrated models will provide more accurate, reliable, and cost‐effective pavement designs than designs created with global calibrations.
17. Key Words Mechanistic‐empirical pavement design, AASHTO, calibration, validation, MEPDG, DARWin‐ME, Long Term Pavement Performance, LTPP, hot‐mix asphalt, concrete, rehabilitation, performance
18. Distribution Statement This document is available to the US public through the National Technical Information Service, Springfield, Virginia, 22161.
23. Registrant's Seal Unclassified.
19. Security Classification
Unclassified.
20. Security Classification
Unclassified.
21. No. of Pages
208
22. Price
* SI is the symbol for the International System of Units. Appropriate rounding should be made to comply with Section 4 of
ASTM E380. (Revised March 2003)
v
Contents
Executive Summary ................................................................................................................. 1
Chapter 1. Introduction ............................................................................................................ 3
Background ................................................................................................................................... 3
Overview of MEPDG Implementation Efforts in Arizona.............................................................. 4
Project Objectives ......................................................................................................................... 4
Report Overview ........................................................................................................................... 5
Chapter 2. Framework for MEPDG Performance Model Verification and Local Calibration ....... 7
Step 1: Select Hierarchical input Levels ........................................................................................ 8
Steps 2 and 3: Determine Experimental Design and Sampling Needs ....................................... 10
Steps 4 and 5: Select HMA and JPCP Projects to Populate Sampling Template ......................... 14
Step 6: Conduct Field and Forensic Investigations ..................................................................... 50
Steps 7 Through 10: Assess Local Bias and Standard Error of the Estimate from Global
Calibration Factors, and Eliminate/Reduce Standard Error of the Estimate and
Local Bias of Distress Prediction Models .............................................................................. 50
Step 11: Interpret Results and Determine Adequacy of ADOT MEPDG Locally
Calibrated Models ................................................................................................................ 54
Chapter 3. Verification, Local Recalibration, and Validation of MEPDG Models ...................... 57
New HMA and HMA Overlaid HMA Alligator Cracking ............................................................... 57
New HMA and HMA Overlaid HMA Rutting ............................................................................... 64
Transverse Low‐Temperature Cracking ...................................................................................... 69
New HMA and HMA Overlaid HMA Smoothness (IRI) ................................................................ 74
New and Reconstructed JPCP Transverse Cracking .................................................................... 79
New JPCP Transverse Joint Faulting ........................................................................................... 86
New JPCP Smoothness (IRI) ........................................................................................................ 90
ARFC/JPCP Composite Pavement Models .................................................................................. 95
Chapter 4. Sensitivity Analysis ............................................................................................. 107
New HMA Sensitivity Analysis Results ...................................................................................... 108
New JPCP Sensitivity Analysis Results ...................................................................................... 125
Chapter 5. Validation of Locally Calibrated DARWin‐ME New HMA and JPCP Design
Procedures for Arizona ..................................................................................................... 139
New HMA Pavement Design Comparison ................................................................................ 139
New JPCP Design Comparison .................................................................................................. 144
Chapter 6. Practical Methodology to Recalibrate DARWin‐ME ............................................. 149
Step 1: Select Hierarchical input Levels .................................................................................... 150
Steps 2 and 3: Develop Experimental Design and Sampling Needs ......................................... 151
Steps 4 and 5: Select HMA and JPCP Pavement Projects to Populate Sampling
Template, and Extract and Evaluate Distress and Project Data ......................................... 153
Step 6: Conduct Field and Forensic Investigations ................................................................... 153
Step 7: Characterize Bias and Goodness of Fit of DARWin‐ME Global Models
Used Under Arizona Conditions ......................................................................................... 153
vi
Step 8: Perform Local Calibration to Reduce Bias and Improve Goodness of Fit ..................... 156
Steps 9 Through 11: Assess and Reduce Standard Error, and Interpret Results and
Determine Adequacy of Locally Calibrated DARWin‐ME Models ...................................... 157
Chapter 7. Arizona Input Defaults and Calibration Coefficients for MEPDG .......................... 159
Default Level 3 Inputs ............................................................................................................... 159
DARWin‐ME Arizona Local Calibration Coefficients ................................................................. 165
Checklist of All Inputs ............................................................................................................... 168
Chapter 8. Summary and Recommendations ........................................................................ 173
References ........................................................................................................................... 177
Appendix: DARWin‐ME HMA Pavement and JPCP Performance Prediction Models .............. 179
vii
List of Figures
Figure 1. Construction/Rehabilitation Year of Projects Selected for the ADOT MEPDG
Calibration/Validation Database ................................................................................ 17
Figure 2. Arizona Locations of Projects Selected for the ADOT MEPDG Calibration/
Validation Database ................................................................................................... 17
Figure 3. Highway Functional Class of Projects Selected for the ADOT MEPDG
Calibration/Validation Database ................................................................................ 18
Figure 4. Volumetric (Not Gravimetric) Binder Content of Projects Selected for the
ADOT MEPDG Calibration/Validation Database ......................................................... 18
Figure 5. In‐Place Air Voids of Projects Selected for the ADOT MEPDG Calibration/
Validation Database ................................................................................................... 19
Figure 6. PCC CTE (from LTPP) of Projects Selected for the ADOT MEPDG Calibration/
Validation Database ................................................................................................... 19
Figure 7. PCC 28‐Day Compressive Strength of Projects Selected for the ADOT
MEPDG Calibration/Validation Database ................................................................... 20
Figure 8. PCC 28‐Day Elastic Modulus of Projects Selected for the ADOT MEPDG
Calibration/Validation Database ................................................................................ 20
Figure 9. Aggregate Base and Subgrade Liquid Limit of Projects Selected for the
ADOT MEPDG Calibration/Validation Database ......................................................... 21
Figure 10. Aggregate Base and Subgrade Plasticity Index of Projects Selected for the
ADOT MEPDG Calibration/Validation Database ......................................................... 22
Figure 11. Aggregate Base and Subgrade Maximum Dry Density of Projects Selected
for the ADOT MEPDG Calibration/Validation Database ............................................. 23
Figure 12. Subgrade Type (Coarse‐ or Fine‐Grained) of Projects Selected for the
ADOT MEPDG Calibration/Validation Database ......................................................... 23
Figure 13. New HMA Pavement Thickness of Projects Selected for the ADOT MEPDG
Calibration/Validation Database ................................................................................ 24
Figure 14. Existing HMA Thickness for HMA Overlaid HMA Pavement Projects Selected
for the ADOT MEPDG Calibration/Validation Database. ............................................ 24
Figure 15. HMA Overlay Thickness of HMA Overlaid HMA Pavement Projects Selected
for the ADOT MEPDG Calibration/Validation Database. ............................................ 25
Figure 16. Relationship Between ADOT and LTPP Rutting Measurements. ................................ 27
Figure 17. Measured Alligator Cracking in LTPP and PMS Sections of New HMA and
HMA Overlaid HMA Projects. ..................................................................................... 29
Figure 18. Measured HMA Transverse Cracking in LTPP and PMS Sections of New HMA
and HMA Overlaid HMA Projects. .............................................................................. 29
Figure 19. Measured Rutting in LTPP and PMS Sections of New HMA and
HMA Overlaid HMA Projects ...................................................................................... 30
Figure 20. Measured AC IRI in LTPP and PMS Sections of New HMA and
HMA Overlaid HMA Projects ...................................................................................... 30
viii
Figure 21. Measured JPCP Transverse Cracking in LTPP and PMS Sections of
New JPCP Projects ...................................................................................................... 31
Figure 22. Measured JPCP Transverse Joint Faulting in LTPP and PMS Sections of
New JPCP Projects ...................................................................................................... 31
Figure 23. Measured JPCP IRI of LTPP and PMS Sections of New JPCP Projects ......................... 32
Figure 24. Progression of Measured IRI with Age in ADOT PMS 03‐31....................................... 33
Figure 25. Progression of Measured Rutting with Age in ADOT LTPP 0509. ............................... 33
Figure 26. Progression of Measured HMA Transverse Cracking with Age in
ADOT LTPP 0113 ......................................................................................................... 34
Figure 27. Progression of Measured JPCP Transverse Cracking with Age in
ADOT LTPP 0217 ......................................................................................................... 34
Figure 28. Location of Selected New HMA and HMA Overlaid HMA Pavement Projects. .......... 38
Figure 29. Location of Selected New Bare JPCP Projects. ........................................................... 39
Figure 30. Location of Selected New ARFC‐surfaced JPCP Composite Projects. ......................... 40
Figure 31. Initial Verification of the HMA Alligator Fatigue Cracking Models with
Global Coefficients Using Arizona Performance Data. ............................................... 57
Figure 32. Measured and Predicted Alligator Cracking Versus Cumulated Fatigue
Damage for the Locally Calibrated Alligator Cracking Submodels ............................. 62
Figure 33. Predicted and Measured Alligator (Fatigue) Cracking for Arizona I‐8 HMA
(LTPP 4_0501) Pavement ........................................................................................... 63
Figure 34. Predicted and Measured Alligator (Fatigue) Cracking for Arizona I‐8 HMA
(LTPP 4_050) Overlaid Pavement ............................................................................... 63
Figure 35. Predicted and Total Rutting Using Global Coefficients and Arizona HMA
Pavement Performance Data ..................................................................................... 64
Figure 36. Measured and Predicted Total Rutting for New HMA and HMA Overlaid
HMA Pavements for the Locally Calibrated Rutting Submodels (Using
100 Percent of Projects) ............................................................................................. 67
Figure 37. High Variation of Measured Rutting for LTPP Project 4_0162 ................................... 68
Figure 38. Measured and Predicted (Global and Arizona Local Calibrated Models) Total
Rutting for Arizona LTPP Project 4_0113 ................................................................... 68
Figure 39. Transverse Cracking in Nonfreeze Areas of Phoenix and Tucson .............................. 69
Figure 40. Monthly Temperatures Recorded at the SPS‐1 Site Near Kingman, Arizona ............. 70
Figure 41. Measured Transverse Cracking at the SPS‐1 Site After 10 Years of Service .............. 70
Figure 42. Predicted and Measured Transverse Cracking Using Global Coefficients
in SPS‐1 HMA Pavement Sections .............................................................................. 71
Figure 43. Predicted (Calibration K = 100) Versus Measured Transverse Cracking
in SPS‐1 HMA Pavement Sections .............................................................................. 72
Figure 44. MEPDG Predicted Transverse Cracking at Various Arizona Locations ....................... 74
Figure 45. Predicted Versus Measured IRI Using Global Coefficients andArizona
HMA Pavement Performance Data ............................................................................ 75
Figure 46. Measured and Predicted HMA IRI for New HMA and HMA Overlaid HMA
Pavements for the Locally Calibrated IRI Model (Using 100 Percent of Projects). .... 77
ix
Figure 47. Measured and Predicted IRI for HMA LTPP Project 4_0260 ...................................... 78
Figure 48. Measured and Predicted (Global and Arizona Local Calibrated Models)
Total IRI for Arizona LTPP Project 4_0122 .................................................................. 78
Figure 49. Predicted Versus Measured Percent Slabs Cracked for Arizona JPCP with
Global Calibration Coefficients ................................................................................... 79
Figure 50. Arizona JPCP with HMA Base Course Measured and Predicted Percent
Slabs Transversely Cracked ........................................................................................ 81
Figure 51. Arizona JPCP with Unbound Aggregate Base Course Measured and Predicted
Percent Slabs Transversely Cracked ........................................................................... 81
Figure 52. Measured and Predicted JPCP Transverse Cracking Using MEPDG Global
Models for JPCP Over Lean Concrete Base ................................................................ 82
Figure 53. Predicted Versus Measured Percent Slabs Transverse Cracked for
Arizona JPCP with All Types of Base Courses ............................................................. 83
Figure 54. Measured and Predicted JPCP Transverse Cracking Versus
Cumulative Fatigue Damage for Local Arizona Calibration Coefficients
(Using 100 Percent of Projects) .................................................................................. 84
Figure 55. Predicted (Using Local Calibration Factors and Input Recommendations) and
Measured Transverse Cracking for JPCP 4_0217 (SPS‐2) with
Lean Concrete Base .................................................................................................... 85
Figure 56. Predicted (Using Local Calibration Factors and Input Recommendations)
and Measured Transverse Cracking for JPCP 4_0214 (SPS‐2) with
Untreated Aggregate Base ......................................................................................... 85
Figure 57. Predicted (Using Local Calibration Factors and Input Recommendations) and
Measured Transverse Cracking for JPCP 4_0224 (SPS‐2) with HMA Base. ................ 86
Figure 58. Predicted Versus Measured Joint Faulting for JPCP with Global
Calibration Coefficients .............................................................................................. 87
Figure 59. Measured and Predicted Transverse Joint Faulting for Locally Calibrated
Faulting Submodels (Using 100 Percent of Projects) ................................................. 89
Figure 60. Measured and Predicted Joint Faulting Using the Arizona Calibrated Model
for JPCP Section 4_0265 on I‐10 ................................................................................. 90
Figure 61. Predicted JPCP IRI Versus Measured Arizona JPCP with Global
Calibration Coefficients .............................................................................................. 91
Figure 62. Measured and Predicted JPCP IRI (Using 100 Percent of Projects) ............................ 93
Figure 63. Predicted and Measured JPCP IRI for Arizona Section 04_0262 ................................ 94
Figure 64. Predicted and Measured JPCP IRI for Arizona Section 04_0267 ................................ 94
Figure 65. Measured Rutting of ARFC (0.75‐inch) for Arizona Composite Pavements ............... 98
Figure 66. Measured Versus Predicted Rutting for ARFC Surfacing Over JPCP ........................... 98
Figure 67. Percentage of Slabs Cracked Versus JPCP Slab Fatigue Damage in
JPCP LTPP and PMS Sections and in Composite Pavement ...................................... 101
Figure 68. MEPDG Predicted Versus Measured IRI for Composite Pavement
Project (2 to 16 Years) .............................................................................................. 102
x
Figure 69. CRCP Observed Punchouts Versus Fatigue Damage for All CRCP Used in the
2007 National Calibration......................................................................................... 105
Figure 70. Effect of AADTT on HMA Alligator Fatigue Cracking ................................................ 109
Figure 71. Effect of AADTT on HMA Total Rutting .................................................................... 109
Figure 72. Effect of Initial AADTT on HMA IRI ........................................................................... 110
Figure 73. Effect of HMA Thickness on HMA Alligator Fatigue Cracking .................................. 111
Figure 74. Effect of HMA Thickness on Total Rutting ................................................................ 111
Figure 75. Effect of HMA Thickness on IRI ................................................................................ 112
Figure 76. Effect of Binder Content on HMA Alligator Cracking ............................................... 113
Figure 77. Effect of Binder Content on HMA Total Rutting ....................................................... 113
Figure 78. Effect of Binder Content on HMA IRI ....................................................................... 114
Figure 79. Effect of Air Voids on HMA Alligator Cracking ......................................................... 115
Figure 80. Effect of Air Voids on HMA Total Rutting ................................................................. 115
Figure 81. Effect of Air Voids on HMA IRI .................................................................................. 116
Figure 82. Effect of PG Binder on HMA Alligator Cracking ........................................................ 117
Figure 83. Effect of PG Binder on HMA Total Rutting ............................................................... 117
Figure 84. Effect of PG Binder on HMA IRI ................................................................................ 118
Figure 85. Effect of Subgrade Type on HMA Alligator Cracking ................................................ 119
Figure 86. Effect of Subgrade Type on HMA Total Rutting ....................................................... 119
Figure 87. Effect of Subgrade Type on HMA IRI ........................................................................ 120
Figure 88. Effect of Climate on HMA Alligator Cracking ............................................................ 121
Figure 89. Effect of Climate on HMA Total Rutting ................................................................... 121
Figure 90. Effect of Climate on HMA IRI .................................................................................... 122
Figure 91. Effect of Reliability on HMA Alligator Cracking ........................................................ 123
Figure 92. Effect of Reliability on HMA Total Rutting ............................................................... 124
Figure 93. Effect of Reliability on HMA IRI ................................................................................ 124
Figure 94. Effect of AADTT on JPCP Transverse Cracking .......................................................... 125
Figure 95. Effect of AADTT on JPCP Mean Joint Faulting .......................................................... 126
Figure 96. Effect of Initial AADTT on JPCP IRI ............................................................................ 126
Figure 97. Effect of PCC Thickness on JPCP Transverse Cracking .............................................. 127
Figure 98. Effect of PCC Thickness on JPCP Mean Joint Faulting .............................................. 128
Figure 99. Effect of PCC Thickness on JPCP IRI .......................................................................... 128
Figure 100. Effect of CTE on JPCP Transverse Cracking ............................................................... 129
Figure 101. Effect of CTE on JPCP Mean Joint Faulting ............................................................... 130
Figure 102. Effect of CTE on JPCP IRI ........................................................................................... 130
Figure 103. Effect of Shoulder Type on JPCP Transverse Cracking ............................................. 131
Figure 104. Effect of Shoulder Type on JPCP Mean Joint Faulting .............................................. 132
Figure 105. Effect of Shoulder Type on JPCP IRI .......................................................................... 132
Figure 106. Effect of Subgrade Type on JPCP Transverse Cracking ............................................. 133
Figure 107. Effect of Subgrade Type on JPCP Mean Joint Faulting ............................................. 134
Figure 108. Effect of Subgrade Type on JPCP IRI ......................................................................... 134
Figure 109. Effect of Climate on JPCP Transverse Cracking ........................................................ 135
xi
Figure 110. Effect of Climate on JPCP Mean Joint Faulting ......................................................... 136
Figure 111. Effect of Climate on JPCP IRI .................................................................................... 136
Figure 112. Effect of Reliability on JPCP Transverse Cracking ..................................................... 137
Figure 113. Effect of Reliability on JPCP Mean Joint Faulting ..................................................... 138
Figure 114. Effect of Reliability on JPCP IRI ................................................................................. 138
Figure 115. New HMA Project Trial Design Structure ................................................................. 140
Figure 116. Predicted Alligator Cracking for the New HMA Trial Design Using DARWin‐ME ..... 142
Figure 117. Predicted Rutting for the New HMA Trial Design Using DARWin‐ME ...................... 143
Figure 118. Predicted IRI for the New HMA Trial Design Using DARWin‐ME ............................. 143
Figure 119. New JPCP Project Trial Design Structure .................................................................. 144
Figure 120. Predicted Transverse Cracking for the New JPCP Trial Design
Using DARWin‐ME .................................................................................................... 147
Figure 121 Predicted Faulting for the New JPCP Trial Design Using DARWin‐ME ..................... 148
Figure 122. Predicted IRI for the New JPCP Trial Design Using DARWin‐ME .............................. 148
xii
List of Tables
Table 1. Recommended MEPDG Hierarchical Input Levels for New HMA and HMA
Overlaid Pavements .................................................................................................................. 9
Table 2. Recommended MEPDG Hierarchical Input Levels for New JPCP .............................................. 9
Table 3. ADOT Pavement Design and Construction Practices .............................................................. 11
Table 4. Sampling Template for New and Rehabilitated HMA Pavements .......................................... 12
Table 5. Sampling Template for New and Rehabilitated JPCP .............................................................. 13
Table 6. Estimated Number of Pavement Projects Required for MEPDG Validation and
Local Calibration ...................................................................................................................... 14
Table 7. Sources for the ADOT MEPDG Calibration and Validation Database ...................................... 16
Table 8. Distress and IRI Data Sources for Local Calibration and Validation. ....................................... 26
Table 9. Comparison of Distress and IRI Values with Design Criteria or Threshold Values .................. 32
Table 10. Breakdown of Selected Projects for Validation and Local Calibration .................................... 36
Table 11. Populated Sampling Template for New and Rehabilitated HMA‐Surfaced Pavements .......... 36
Table 12. Populated Sampling Template for New and Rehabilitated JPCP ............................................ 37
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration ......................... 41
Table 14. National Calibration Under NCHRP 1‐40D New HMA Pavement and New JPCP Model
Statistics .................................................................................................................................. 51
Table 15. Sampling Template for the 10 Percent Validation/90 Percent Calibration Databases
for New and Rehabilitated HMA‐Surfaced Pavements ........................................................... 55
Table 16. Simplified Sampling Template for the Validation and Local Calibration of New JPCP ............ 56
Table 17. MEPDG Global Models Evaluated for Arizona Local Conditions ............................................. 58
Table 18. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics for
Alligator Cracking Submodels Using 90 Percent of Projects ................................................... 60
Table 19. Goodness of Fit Statistics for Validating the Preliminary 90 Percent Locally Calibrated
Alligator Cracking Submodels with 10 Percent of Projects. .................................................... 60
Table 20. Goodness of Fit and Bias Test Statistics for Final Locally Calibrated Alligator Cracking
Model (Based on 100 Percent of All Projects). ........................................................................ 61
Table 21. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics for Total
Rutting Submodels Using 90 Percent of Projects. ................................................................... 66
Table 22. Goodness of Fit Statistics for Validating the Preliminary 90 Percent Locally Calibrated
Total Rutting Submodels with 10 Percent of Projects. ............................................................ 66
Table 23. Goodness of Fit and Bias Test Statistics for Final Locally Calibrated Rutting Model
(Based on 100 Percent of All Projects) .................................................................................... 67
Table 24. Measured and Predicted HMA Transverse Cracking at the SPS‐1 Site ................................... 72
Table 25. MEPDG Prediction of HMA Transverse Cracking for Arizona Climates
(Calibration K = 50) .................................................................................................................. 73
Table 26. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics for
HMA IRI Model Using 90 Percent of Projects .......................................................................... 76
Table 27. Goodness of Fit Statistics for Validating the 90 Percent Preliminary Locally
Calibrated HMA IRI Model with 10 Percent of Projects .......................................................... 76
xiii
Table 28. Goodness of Fit and Bias Test Statistics for Final Locally Calibrated HMA IRI Model
(Based on 100 Percent of All Projects) .................................................................................... 77
Table 29. Goodness of Fit and Bias Test Statistics for Globally Calibrated JPCP Transverse
Cracking Model (Based on 100 Percent of All Projects) .......................................................... 83
Table 30. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics for
Transverse Joint Faulting Model Using 90 Percent of Projects ............................................... 88
Table 31. Goodness of Fit Statistics for Validating the 90 Percent Preliminary Locally Calibrated
Transverse Joint Faulting Model with 10 Percent of the Data ................................................ 88
Table 32. Goodness of Fit and Bias Test Statistics for Final Locally Calibrated Transverse
Joint Faulting Model (Based on 100 Percent of All Projects) .................................................. 89
Table 33. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics for
JPCP IRI Model Using 90 Percent of Selected Calibration Projects ......................................... 92
Table 34. Goodness of Fit Statistics for Validating the 90 Percent Preliminary Locally
Calibrated JPCP IRI Model with the 10 Percent Model ........................................................... 92
Table 35. Goodness of Fit and Bias Test Statistics for Final Arizona Calibrated JPCP IRI Model
(Based on 100 Percent of All Projects) .................................................................................... 93
Table 36. Arizona Composite Pavement Rutting and IRI Calibration Data ............................................. 97
Table 37. Fatigue Damage and Cracking Analysis for ARFC/JPCP Composite Pavements ...................... 99
Table 38. Measured and Predicted Performance of Loop 101 and I‐10 CRCP Sections ....................... 104
Table 39. Key Inputs in New or Reconstructed HMA Pavement Sensitivity Analysis ........................... 107
Table 40. Key Inputs in New or Reconstructed JPCP Sensitivity Analysis ............................................. 108
Table 41. Effect of Climate on Predicted Distress and IRI ..................................................................... 122
Table 42. Key Inputs for New HMA Pavement Project Design Comparison ....................................... 1410
Table 43. Trial Design Results for New HMA Pavement Projects ....................................................... 1421
Table 44. Key Inputs for New JPCP Project Design ............................................................................... 146
Table 45. Trial Design Results for New JPCP Project Design ................................................................. 147
Table 46. Recommended DARWin‐ME Hierarchical Input Levels for New HMA and
HMA Overlaid Pavements ..................................................................................................... 150
Table 47. Recommended DARWin‐ME Hierarchical Input Levels for New JPCP................................... 151
Table 48. Sampling Template for New and Rehabilitated HMA Pavements ........................................ 152
Table 49. Estimated Number of Pavement Projects Required for Validation and Local
Calibration ............................................................................................................................. 152
Table 50. National Calibration under NCHRP 1‐40D of New HMA Pavement and New
JPCP Model Statistics ............................................................................................................. 155
Table 51. Default Level 3 Inputs for HMA Materials ............................................................................. 159
Table 52. Default Level 3 PCC Compressive Inputs ............................................................................... 160
Table 53. Default Level 3 PCC Flexural Strength Inputs ........................................................................ 160
Table 54. Default Level 3 PCC Elastic Modulus Inputs .......................................................................... 161
Table 55. Default Level 3 PCC CTE Inputs .............................................................................................. 161
Table 56. Default Level 3 PCC Mixture Constituents Inputs ................................................................. 161
Table 57. Default Level 3 Inputs for Cement Aggregate Mixture Materials ......................................... 162
Table 58. Default Level 3 Inputs for Unbound Aggregate Materials .................................................... 163
xiv
Table 59. Default Level 3 Inputs for Subgrade Materials Mr at Optimum Moisture Content .............. 164
Table 60. DARWin‐ME Local Calibration Coefficients for New HMA and HMA Overlaid
HMA Pavement ..................................................................................................................... 166
Table 61. DARWin‐ME Local Calibration Coefficients for New JPCP ..................................................... 167
Table 62. DARWin‐ME Local Calibration Coefficients for New CRCP .................................................... 167
Table 63. Checklist of Flexible Pavement Inputs ................................................................................... 168
Table 64. Checklist of Rigid Pavement Inputs ....................................................................................... 170
Table 65. Comparison of MEPDG Global and Arizona‐Specific Model Goodness of Fit Statistics ........ 175
Table 66. HMA Pavement Local Calibration Coefficients ...................................................................... 196
Table 67. JPCP Local Calibration Coefficients ........................................................................................ 196
xv
List of Acronyms
AADTT ........................................................................................... average annual daily truck traffic
AASHTO ................................. American Association of State Highway and Transportation Officials
AC ............................................................................................................................ asphalt concrete
ADOT .................................................................................... Arizona Department of Transportation
AR ............................................................................................................................... asphalt rubber
ARFC ................................................................................................... asphalt rubber friction course
ASU .............................................................................................................. Arizona State University
CPR ................................................................................................... concrete pavement restoration
CRCP ............................................................................ continuously reinforced concrete pavement
CTE ................................................................................................. coefficient of thermal expansion
FHWA ............................................................................................. Federal Highway Administration
GPS ............................................................................................................ global positioning system
HMA .......................................................................................................................... hot‐mix asphalt
IRI ...................................................................................................... International Roughness Index
JPCP ............................................................................................... jointed plain concrete pavement
JTFP .................................................................................................. Joint Task Force on Pavements
LTPP ............................................................................................Long‐Term Pavement Performance
MEPDG .................................................................... Mechanistic Empirical Pavement Design Guide
NCDC ................................................................................................... National Climate Data Center
NCHRP ................................................................ National Cooperative Highway Research Program
NRCS ................................................................................... Natural Resources Conservation Service
PCC ........................................................................................................... portland cement concrete
PG ........................................................................................................................ performance grade
PMS ............................................................................................... Pavement Management Services
QA .......................................................................................................................... quality assurance
SEE ..................................................................................................... standard error of the estimate
SHA ...................................................................................................... state highway administration
SPR ....................................................................................................... State Planning and Research
SPS ............................................................................................................... specific pavement study
SSURGO ......................................................................................... Soil Survey Geographic database
USDA ................................................................................. United States Department of Agriculture
xvi
1
EXECUTIVE SUMMARY
This report documents the efforts of the Arizona Department of Transportation (ADOT) to implement
the American Association of State Highway and Transportation Officials (AASHTO) DARWin‐ME
pavement design and rehabilitation guide. As part of this implementation, the research team also
prepared a practical stand‐alone user’s guide that provides guidance for obtaining inputs, conducting
design, and establishing the recommended pavement design.
Adopting AASHTO DARWin‐ME will result in more accurate, reliable, and cost‐effective ADOT pavement
and rehabilitation designs. The objectives of ADOT’s implementation efforts were to:
Ensure that all design inputs were proper and were tailored to Arizona conditions and resources.
Ensure that the distress and International Roughness Index (IRI) prediction models were
unbiased (i.e., did not consistently over‐ or underpredict).
Reduce the prediction error of the distress and IRI models. (Each model’s error is used in the
reliability design and thus affects the resulting design.)
Provide a user’s guide and training for ADOT designers and consultants.
ADOT identified the following desired pavement design applications in DARWin‐ME for implementation:
Flexible (hot‐mix asphalt [HMA]) pavements.
Composite pavements (thin asphalt rubber friction course over jointed plain concrete pavement
[JPCP] and continuously reinforced concrete pavement [CRCP]).
JPCP and CRCP (bare designs).
HMA overlays of flexible pavement.
To achieve these objectives, the researchers used Microsoft Excel to create an ADOT pavement
database that comprised 180 pavement sections and inputs for the DARWin‐ME. The database included:
The four pavement design applications. The project team obtained pavement sections from the
Long‐Term Pavement Performance (LTPP) experiment, previous field research studies, and other
projects included in the pavement management database (SODA).
Design, materials, and construction inputs for each section, obtained from the LTPP database
and ADOT files.
Past traffic loading information, also obtained from the LTPP database and ADOT files.
Measured performance data over the life of the pavement, obtained from the LTPP database,
ADOT files and videos, and manual field surveys conducted by the research team.
2
After reviewing and analyzing all DARWin‐ME inputs for their importance and sensitivity, the researchers
identified several key inputs to more accurately define them and to develop Arizona defaults. These
inputs (listed below) were based on data from the ADOT pavement database established for this study:
Default values for all design, traffic, materials, and construction inputs that are specific to
Arizona.
Subgrade resilient modulus (Arizona‐specific soils based on backcalculation).
Time of full slab and base friction for various base courses.
Performance criteria and design reliability levels for all classes of highways.
Traffic inputs (gathered from work accomplished under the ADOT project, Development of a
Traffic Data Input System in Arizona for the MEPDG (Darter et al. 2010)).
Weather station data correction.
The researchers then applied a formal verification, calibration, and final validation process of the
distress and IRI models to each distress and IRI model in DARWin‐ME, using Arizona data from the 180
sections. Verification involved testing the model predictions with global coefficients but using only
Arizona performance data. If the model showed bias (over‐ or underprediction overall), it was identified
for recalibration. Recalibration involved deriving new local coefficients for each model using the Arizona
performance data that removed the bias and reduced the prediction error. Validation involved a further
independent check of the model using 10 percent of the data withheld from the recalibration effort.
This process removed bias (consistent over‐ or underprediction) and improved accuracy of model
prediction. In most cases, this effort improved distress and IRI model accuracy and removed the over‐
and underprediction of bias. For example, the goodness of fit for the HMA IRI model improved from R2 =
30 percent with global coefficients to 80 percent with Arizona‐specific coefficients. The standard error of
IRI went from 19 to 8 inch/mi. The improvement in these models was notable, indicating that the
Arizona‐calibrated models will provide much more accurate, reliable, and cost‐effective designs than
models using global calibrations.
The researchers also prepared design comparisons and sensitivity studies that helped to establish
confidence in the pavement design results from the Mechanistic‐Empirical Pavement Design Guide.
3
CHAPTER 1. INTRODUCTION
BACKGROUND
In 1998 the American Association of State Highway and Transportation Officials (AASHTO) Joint Task
Force on Pavements (JTFP) initiated National Cooperative Highway Research Program (NCHRP) Project
1‐37A, Development of the Guide for Design of New and Rehabilitated Pavement Structures: Phase II, to
develop a mechanistic‐based design procedure for new and rehabilitated pavement structures. This
project resulted in the Mechanistic‐Empirical Pavement Design Guide (MEPDG) in 2004. The project
deliverables consisted of a guide for mechanistic‐empirical pavement design and analysis, companion
software and a software user’s manual, and implementation and training materials.
Following completion of NCHRP Project 1‐37A, the AASHTO JTFP developed a plan for state highway
agencies (SHAs) to adopt and implement the MEPDG. Key components of the plan, which was
implemented under NCHRP Projects 01‐40A, 01‐40B, 01‐40D, and 20‐07, included developing guides and
software tools such as the AASHTO MEPDG (Interim Edition), AASHTO MEPDG Local Calibration Guide,
and DARWin‐ME (the professional/commercial‐grade software tool based on the MEPDG design and
analysis principles). These products collectively allow users to analyze and perform new and
rehabilitated pavement design and forensic analysis such as:
Modeling truck traffic growth, truck class distribution, various volume adjustments, axles per
truck type, and axle load distribution.
Modeling temperature and moisture conditions within the pavement structure and subgrade.
Characterizing time‐ and climate‐dependent paving materials properties.
Computing critical pavement responses (such as stresses, strains, and deflections) to applied
traffic loads using layer elastic and finite‐element‐based analysis techniques.
Predicting pavement damage and distress using transfer models and smoothness.
Designing a variety of flexible, rigid, and rehabilitated pavement structures to meet selected
performance criteria for a specified level of reliability.
The MEPDG distress and smoothness models were calibrated using data from hundreds of flexible, rigid,
and rehabilitated pavement projects included in the Federal Highway Administration (FHWA) Long‐Term
Pavement Performance (LTPP) program as well as additional state and other research sections. These
projects were located throughout the United States, including Arizona. Calibration here indicates that
various model coefficients were derived so that the cracking, rutting, faulting, and smoothness models
predict as closely as possible similar distresses that exist in the field for hundreds of projects.
4
OVERVIEW OF MEPDG IMPLEMENTATION EFFORTS IN ARIZONA
The Arizona Department of Transportation (ADOT) and other U.S. SHAs initiated efforts to adopt the
MEPDG as the standard procedure for designing new and rehabilitated pavements. ADOT’s MEPDG
implementation plans began in 2008 and have focused on:
Characterizing asphalt binder, hot‐mix asphalt (HMA), granular base course, and subgrade soils
to provide inputs to the MEPDG.
Characterizing traffic loadings, volumes, vehicle distribution, lane distribution, lateral
distribution, and other inputs under Darter et al. 2010.
Identifying the desired MEPDG applications, which included flexible (HMA) pavements, jointed
plain concrete pavement (JPCP), composite pavements (thin asphalt rubber friction [ARFC]
course over JPCP and continuously reinforced concrete pavement [CRCP]), HMA overlays of
flexible pavement, concrete pavement restoration (CPR), and JPCP overlays.
Collecting and assembling MEPDG input data from 180 LTPP and pavement management system
sections of flexible, rigid, composite, and rehabilitated pavements for use in local calibration.
Calibrating the MEPDG (including the accompanying analysis and design software) for select
applications in Arizona conditions (such as highway pavement structures, materials, traffic, and
climate).
Validating the calibrated MEPDG, including the accompanying analysis and design software, for
Arizona conditions.
This report provides details of all activities, results, and recommendations needed for ADOT to
implement the MEPDG.
PROJECT OBJECTIVES
This project’s objectives were to:
Evaluate the applicability of the MEPDG for Arizona conditions.
Calibrate the MEPDG, including the accompanying analysis and design software, for Arizona
conditions such as highway pavement structures, materials, traffic, and climate.
Validate the calibrated MEPDG, including the accompanying analysis and design software, for
Arizona conditions.
Finalize a comprehensive MEPDG and associated analysis and design software user or practice
manual for Arizona.
5
Develop a practical methodology and tools for use in periodically recalibrating the Arizona‐
calibrated MEPDG.
Deliver a populated and implementable calibrated Arizona MEPDG along with associated
software based on this work, including all material types and pavement structures commonly
used in Arizona.
Deliver an ADOT MEPDG user’s manual.
Train select ADOT staff in the use and structure of the Arizona MEPDG.
REPORT OVERVIEW
The remainder of this report is organized into eight chapters. Chapter 2 provides the overall framework
for calibrating the MEPDG models for local Arizona conditions and practices. Chapter 3 describes efforts
to verify, calibrate, and validate the MEPDG performance models and design procedure for Arizona.
Chapter 4 includes a comprehensive sensitivity analysis performed to verify reasonableness of the ADOT
locally calibrated MEPDG models and design procedure. Chapter 5 presents comparisons of new flexible
and rigid pavement designs using current empirical ADOT pavement design procedures and the ADOT
locally calibrated MEPDG. Chapter 6 describes the procedure and methodology for future recalibration
of the MEPDG performance models and design procedure. Chapter 7 describes the default traffic and
materials data used in Arizona pavement designs as well as the locally calibrated performance model
coefficients for the MEPDG. Chapter 7 also provides input defaults and Arizona calibration coefficients.
Chapter 8 summarizes this study’s findings and provides recommendations for future improvement.
6
7
CHAPTER 2. FRAMEWORK FOR MEPDG PERFORMANCE MODEL
VERIFICATION AND LOCAL CALIBRATION
The AASHTO Guide for the Local Calibration of the Mechanistic‐Empirical Pavement Design Guide (2010),
developed under NCHRP Project 1‐40B, provides an 11‐step roadmap for calibrating the MEPDG
software to local conditions, policies, and materials (AASHTO 2010):
Step 1—Select hierarchical input levels.
Step 2—Develop experimental factorial design and matrix or sampling template.
Step 3—Estimate minimum sample size (number of pavement projects) required for each
distress/International Roughness Index (IRI) prediction model validation and local calibration.
Steps 4—Select projects to populate sampling template.
Step 5—Extract and assess distress and project data.
Step 6—Conduct field and forensic investigations.
Step 7—Assess local bias: validation of global calibration values to local conditions, policies, and
materials.
Step 8—Eliminate local bias of distress and IRI prediction models.
Step 9—Assess the standard error of the estimate.
Step 10—Reduce standard error of the estimate.
Step 11—Interpret results and determine adequacy of ADOT MEPDG locally calibrated models.
This process was adopted with modifications for calibrating the MEPDG for ADOT. This chapter describes
the local calibration framework and evaluates its suitability for local Arizona conditions.
8
STEP 1: SELECT HIERARCHICAL INPUT LEVELS
In the MEPDG, pavement design inputs such as traffic, materials, climate, and rehabilitation are
available at three hierarchical levels (AASHTO 2008):
Level 1: Material input requires laboratory or field testing such as the dynamic modulus (E*)
testing of HMA concrete, coefficient of thermal expansion (CTE) of concrete, or falling weight
deflectometer (FWD) deflection testing. Level 1 inputs for traffic require on‐site measurement
of axle load distribution, truck lane usage, and truck classification. Obtaining Level 1 inputs
requires more resources and time than other levels. Level 1 inputs are typically used for
designing heavily trafficked pavements or wherever there is dire safety or economic
consequences of early failure.
Level 2: Inputs are user‐selected (possibly from an agency database), can be derived from a
limited testing program, or can be estimated through correlations of simpler tests with the more
complicated inputs for the DARWin‐ME. Examples include estimating HMA dynamic modulus
(E*) from binder, aggregate, and mix properties; estimating portland cement concrete (PCC)
elastic moduli from compressive strength tests; or using traffic classification data based on
functional class of highway in the state. This level is used when resources or testing equipment
are not available for tests required for Level 1.
Level 3: Inputs are user‐selected values or typical averages for the region. Examples include
Arizona default unbound materials’ resilient modulus values or default PCC CTE for a given
coarse aggregate type. This level might be used for design where there are minimal
consequences of early needed rehabilitation such as lower volume roads.
The Guide for the Local Calibration of the Mechanistic‐Empirical Pavement Design Guide (2010)
recommends selecting an appropriate hierarchical input level (1, 2, or 3) that is consistent with the
highway agency’s day‐to‐day practices for characterizing pavement inputs for design. Another
consideration for selecting input level is how sensitive the given input is to predicted performance and
thus designs.
Tables 1 and 2 summarize the hierarchical levels for flexible and rigid pavement design inputs. The
recommended level is derived from both sensitivity analysis outcomes and ADOT practices for
characterizing pavement materials properties, subgrade properties, and traffic.
9
Table 1. Recommended MEPDG Hierarchical Input Levels for New HMA and HMA Overlaid Pavements
MEPDG Input Variable Sensitivity to Predicted Distress/Smoothness Recommended
Hierarchical Input Level
Alligator Cracking
Rutting Transverse Cracking
IRI
HMA thickness XXX XX X XX Level 1
HMA coefficient of thermal contraction XX Levels 2 and 3
HMA dynamic modulus XX XXX Level 2
HMA air voids in situ (at placement) XXX XXX XX Level 3
Effective HMA binder content XXX XX XX X Level 3
HMA creep compliance XX XXX XXX Level 1
HMA tensile strength XXX Level 3
Base type/modulus XXX XX Level 2
Base thickness X Level 1
Subgrade type/modulus XX XX Level 1
Groundwater table X X Level 3
Climate XX XX XXX X Level 2
Truck volume XXX XXX Level 2
Truck axle load distribution X X Level 2
Tire load, contact area, and pressures XX XXX Level 3
Truck speed XX XXX Level 3
Truck wander XX XX Level 3
Initial IRI XXX Level 2
X = Small effect on distress/IRI. XX = Moderate effect on distress/IRI. XXX = Large effect on distress/IRI.
Table 2. Recommended MEPDG Hierarchical Input Levels for New JPCP
MEPDG Input Variable Sensitivity to Predicted Distress/Smoothness Recommended
Hierarchical Input Level Faulting Transverse Cracking IRI
PCC thickness XX XXX XX Level 1
PCC modulus of rupture and elasticity
XXX X Levels 2 and 3
PCC CTE XXX XXX XXX Levels 1 and 2
(current data available)
Joint spacing XX XXX XX Level 1
Lane to PCC shoulder long‐term load transfer efficiency
XXX XXX Level 3
Edge support XX XXX XX Level 1
Permanent curl/warp XXX XXX XXX Level 3
Base type XXX XXX X Level 1
Climate XXX XXX XXX Level 2
Subgrade type/modulus X XX X Level 1
Truck axle load distribution X XXX X Level 1
Truck volume XXX XXX XXX Level 1
Tire pressure X Level 3
Truck lateral offset XX XXX XX Level 3
Truck wander XX Level 3
Initial IRI XXX Level 1 or 2
X = Small effect on distress/IRI. XX = Moderate effect on distress/IRI. XXX = Large effect on distress/IRI.
10
STEPS 2 AND 3: DETERMINE EXPERIMENTAL DESIGN AND SAMPLING NEEDS
Experimental Design
In Step 2 of the local calibration effort, the researchers developed a statistical experiment to verify that
the MEPDG global models and procedure were adequate for Arizona local conditions and practices, and
to calibrate the global models if they were inadequate. Meeting these objectives involved defining the
population, determining sampling needs and extent, and establishing the experimental design.
Defining the pavement population required a thorough review of past and current ADOT pavement
design and construction practices and planned innovation regarding future ADOT pavement design and
construction. Issues identified from this review follow:
Past and current ADOT design and construction practices:
o Pavement types (new and rehabilitation).
o Material types.
o Site properties (subgrade, climate, and traffic).
o Design features.
o Types of rehabilitation practices employed.
o Construction practices.
Future ADOT design and construction practices:
o New materials (e.g., warm mix asphalt and polymer modified asphalt).
o Incentive‐based construction (e.g., warranties, design build, and performance‐related
specifications).
o New construction techniques and equipment (e.g., intelligent compaction).
Table 3 summarizes the information gathered about these issues.
11
Table 3. ADOT Pavement Design and Construction Practices
Design/Construction Issues Current/Past ADOT Practices Future ADOT Practices
Pavement types (new and
rehabilitation)
New flexible (asphalt concrete [AC] over granular base).
New JPCP over granular base. Composite thin AC over JPCP.
Same. No significant change in
pavement design types
anticipated.
Material types
AC (conventional and Superpave mixes).
Granular (base) materials (typically
AASHTO Class A‐1 and A‐2).
PCC (typical 28‐day compressive strength
of 5000 psi and CTE of 4.7 inch/° F).
Same. No significant change
anticipated.
Site properties
Subgrade (mostly coarse grained). Pockets
of fine‐grained soils present.
Same. No significant change
anticipated.
Climate (primarily change in elevation). Same. No significant change
anticipated.
Traffic (three main vehicle class
distributions).
Same. No significant change
anticipated.
Design features
JPCP (only): o Joint spacing.
o Shoulder type.
o Slab width.
o Load transfer mechanism.
Same. No significant change
anticipated.
Rehabilitation practices
Existing AC pavements (thin/thick AC
overlays and mill and fill).
Existing JPCP (thin AC overlays).
Same. No significant change
anticipated.
Construction practices
Performance targets (e.g., initial
smoothness, PCC compressive strength,
AC air voids, and AC Marshall strength)
deployed.
Incentive based construction a
possibility.
Based on this information, the researchers defined the pavement population and properties to include
current and past ADOT practices:
New pavement types (AC/granular, JPCP/granular, thin AC/JPCP).
Rehabilitated pavement types (AC/AC, AC/JPCP).
Materials:
o AC (conventional and Superpave mixes).
o Granular (base) materials (typically A‐1 and A‐2).
o PCC (typical 28‐day compressive strength of 5000 psi and CTE of 4.7 inch/° F).
12
Subgrade (mostly coarse‐grained material).
Climate (northern, central, and southern regions of state; low and high elevations).
Traffic (three vehicle class distribution representative of the entire state).
Design features:
o Joint spacing (mostly 15 ft).
o Shoulder type (mostly tied PCC).
o Slab width (12 ft).
o Load transfer mechanism (mostly doweled).
Sampling
Creating a Sampling Template
To collect data required for this analysis, the researchers developed a sampling template representing
key aspects of the ADOT pavement population: current and future ADOT pavement design features,
material types, site conditions, and construction practices. Table 4 shows the sampling template for
flexible pavements, which contained eight cells, while Table 5 indicates 64 cells for JPCP. Note that not
all cells represent current or future ADOT design and construction practices.
Table 4. Sampling Template for New and Rehabilitated HMA Pavements
HMA Thickness (inches)
Granular Base Thickness (inches)*
Subgrade Type
Coarse‐Grained Soils (AASHTO Class A‐1 through A‐3)
Fine‐Grained Soils (AASHTO Class A‐4 through A‐7)
<8 <6 1 2
>6 3 4
>8 <6 5 6
>6 7 8
*All projects had granular base. Note: An equivalent dense‐graded aggregate base was assumed for permeable asphalt‐treated
bases (PATBs).
13
Table 5. Sampling Template for New and Rehabilitated JPCP
PCC Thickness (inches)
Dowel Diameter (inches)
Edge Support*
Subgrade and Base Type
Coarse‐Grained Subgrade Soils (AASHTO Class A‐1 through A‐3)
Fine‐Grained Subgrade Soils (AASHTO Class A‐4 through A‐7)
Lean Concrete Base
Granular or Asphalt‐Treated
Base
Lean Concrete Base
Granular or Asphalt‐Treated
Base
< 10
No dowels None 1 2 3 4
Yes 5 6 7 8
Doweled None 9 10 11 12
Yes 13 14 15 16
> 10
No dowels None 17 18 19 20
Yes 21 22 23 24
Doweled None 25 26 27 28
Yes 29 30 31 32
*Tied PCC and or widened lanes.
Minimum Sample Size Required for Each Distress/IRI Prediction Model Validation and Local Calibration
Under this step, the project team estimated the minimum number of projects required for MEPDG
distress/IRI model validation and local calibration. Information required for this task included model
error (i.e., standard error estimate [SEE]); confidence interval for statistical analysis; and performance
indicators’ threshold value at typical agency design reliability level.
Table 6 summarizes the results from this analysis. The researchers assumed the following parameters
for estimating the minimum number of projects (18 flexible and 21 JPCP):
Design reliability level: 90 percent.
Confidence interval: 90 percent.
SEE for the global MEPDG models as presented in Table 6.
ADOT distress/IRI threshold value as presented in Table 6.
14
Table 6. Estimated Number of Pavement Projects Required for
MEPDG Validation and Local Calibration
Pavement Type
Distress/IRI Distress/IRI Threshold (at 90% Reliability)
SEE
Minimum Number of
Projects Required for Validation and Local Calibration
Minimum Number of Projects
Required for Each Pavement Type
(n)*
New HMA and HMA overlaid HMA
Alligator cracking
20% lane area 5.01% 16
18 Transverse thermal cracking
Crack spacing >100 ft of 630 ft/mi
150 ft/mi** 18
Rutting 0.4 inch 0.107 inch 14
IRI 169 inch/mi 18.9 inch/mi 80
New JPCP
Faulting <0.15 inch 0.033 inch 21
21 Transverse cracking
<10% slabs 4.52% 5
IRI 169 inch/mi 17.1 inch/mi 98
*2
2/
E
Zn
, where 2/Z = 1.601 (for a 90 percent confidence interval), = performance indicator threshold
(design criteria), and E = tolerable bias at 90 percent reliability (1.601*SEE).
**Estimated from other MEPDG implementation projects.
Note: In selecting the overall minimum number of required pavement HMA and JPCP projects, the performance
indicator IRI was excluded because the accuracy of the IRI models depends very much on the accuracy of other
pavement distress predictions. Sampling a vast number of projects to validate the IRI models is therefore
unnecessary if the individual distress prediction models are considered accurate and reasonable.
STEPS 4 AND 5: SELECT HMA AND JPCP PROJECTS TO POPULATE SAMPLING TEMPLATE
Identifying Projects for an ADOT MEPDG Local Calibration/Validation Database
To populate the sampling templates developed in Step 3, the research team selected projects based on
recommendations from the Guide for the Local Calibration of the Mechanistic‐Empirical Pavement
Design Guide (2010):
Projects should be representative of Arizona pavement design and construction practices.
Projects should be representative of typical pavement condition (i.e., poor, moderate, and
good).
Project age should span the range typical of Arizona practice (i.e., newly constructed, older
existing, and rehabilitated).
Projects must be located throughout the state.
15
First, the researchers identified potential projects that met these recommendations. The main sources
of these projects were the LTPP (specifically, LTPP projects located in Arizona); research and forensic
reports containing detailed descriptions of in‐service pavements; and ADOT pavement management
system (PMS) databases. The pavement types in these projects were new HMA pavements (including
HMA pavements subjected to routine or preventive maintenance); HMA overlaid existing HMA
pavement; new JPCP (including JPCP over existing aggregate, asphalt‐treated, and cement‐treated
bases); and composite ARFC‐surfaced new JPCP.
Selecting Projects to Populate the Sampling Template
Next, the project team reviewed the identified projects to determine the availability of sufficiently
detailed input data (such as traffic, materials, and construction) and distress/IRI data in MEPDG‐required
formats. Once these projects were identified, the team:
Extracted and reviewed key MEPDG input data to identify and correct or eliminate anomalies
and outliers.
Extracted and reviewed distress/IRI data to identify and correct or eliminate anomalies and
outliers.
Compared distress/IRI magnitudes to the design threshold values (Table 6).
Assembled the ADOT MEPDG calibration and validation database.
Extract and Review Key MEPDG Input Data for Selected Projects
As shown in Tables 1 and 2, the MEPDG requires several categories of input data. For this
implementation project, pavement data were primarily obtained from the LTPP; ADOT PMS traffic and
inventory data tables; ADOT design and construction reports, and construction quality assurance (QA)
testing databases; and pavement research and forensic examination reports. The researchers obtained
additional data as needed from national pavement, climate, and soils databases, including MEPDG
default pavement‐related data; the National Climatic Data Center (NCDC) database; and the U.S.
Department of Agriculture (USDA) Natural Resources Conservation Service (NRCS) Soil Survey
Geographic (SSURGO) database. Table 7 presents the data sources; Figures 1 through 15 summarize the
data that were extracted and reviewed.
16
Table 7. Sources for the ADOT MEPDG Calibration and Validation Database
Data Type Project Source
LTPP Research/Forensic ADOT Pavement Mgt Services (PMS)
Inventory
INV_GENERAL
INV_AGE
INV_ID
SPS_GENERAL
SPS _ID
Smith et al. 1991
Western Research
Institute (WRI)
reports
dbo_SODAMASTER
Traffic
Darter et al. 2010
Smith et al. 1991
WRI research reports
dbo_SODAMASTER
dbo_SODAMASTER
SHSkdtFactors2007‐2008ver4
SHS2028AADTForecastsver2
SHStrafficProjectionsDataDictionary
Mar2010
Materials
INV_PCC_MIXTURE
INV_PMA_ASPHALT
TST_L05B
TST_PC01
TST_PC02
TST_PC03
TST_PC04
TST_PC08
TST_PC09
TST_SS01_UG01_UG02
TST_SS14_UG14
TST_UG04_SS03
Smith et al. 1991
WRI research reports
Soil Aggregate Tabulation
(Subgrade 12‐06‐10)
Soil Aggregate Tabulation
(Aggregate 12‐06‐10)
Soil Aggregate Tabulation
(Aggregate Base 12‐06‐10)
Soil Aggregate Tabulation
(Aggregate Subbase 12‐06‐10)
Soil Aggregate Tabulation
(Embankment 12‐06‐10)
Soil Aggregate Tabulation
(Borrow 12‐06‐10)
Soil Aggregate Tabulation
(ARACFC 12‐06‐10)
ADOT Materials Specifications
Climate MEPDG NCDC database MEPDG NCDC
database
MEPDG NCDC database
Design
INV_PCC_JOINT
SPS‐1, ‐2, ‐5, ‐6
construction reports
Smith et al. 1991
WRI research reports
ADOT pavement design manual
ADOT projects materials/design
reports
Construction SPS‐1, ‐2, ‐3, ‐5, ‐6, ‐9
construction reports
Smith et al. 1991
WRI research reports
ADOT projects materials/design
reports
17
Figure 1. Construction/Rehabilitation Year of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
Figure 2. Arizona Locations of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
18
Figure 3. Highway Functional Class of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
½-in mix
Base mix
¾-in mix
Figure 4. Volumetric (Not Gravimetric) Binder Content of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
19
Base mix
¾-in mix
½-in mix
Figure 5. In‐Place Air Voids of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
Figure 6. PCC CTE (from LTPP) of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
(Note that the concrete CTE values used for all calibration work
were those corrected values obtained from the FHWA/LTPP lab testing.)
20
Figure 7. PCC 28‐Day Compressive Strength of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
Figure 8. PCC 28‐Day Elastic Modulus of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
21
Aggregate Base
Embankment & Subgrade
Figure 9. Aggregate Base and Subgrade Liquid Limit of Projects Selected
for the ADOT MEPDG Calibration/Validation Database
22
Aggregate Base
Embankment& Subgrade
Figure 10. Aggregate Base and Subgrade Plasticity Index of Projects Selected
for the ADOT MEPDG Calibration/Validation Database
23
Aggregate Base
Embankment& Subgrade
Figure 11. Aggregate Base and Subgrade Maximum Dry Density of Projects
Selected for the ADOT MEPDG Calibration/Validation Database
Figure 12. Subgrade Type (Coarse‐ or Fine‐Grained) of Projects Selected
for the ADOT MEPDG Calibration/Validation Database
24
Figure 13. New HMA Pavement Thickness of Projects Selected for the
ADOT MEPDG Calibration/Validation Database
Figure 14. Existing HMA Thickness for HMA Overlaid HMA Pavement Projects
Selected for the ADOT MEPDG Calibration/Validation Database
25
Figure 15. HMA Overlay Thickness of HMA Overlaid HMA Pavement Projects
Selected for the ADOT MEPDG Calibration/Validation Database
Once assembled, the key inventory, traffic, materials, climate, design, and construction data were
reviewed to identify and correct missing data, outliers, and errors:
The project team removed questionable data and replaced the information with typical values
or with project‐specific information from other sources.
Team members used assembled historical traffic volume data to compute initial truck traffic
volume and future traffic growth rates.
They also used Level 3 typical values developed from ADOT materials and traffic databases for
ADOT and research and forensic projects with inadequate amounts of project‐specific traffic and
materials data.
After the data review, the researchers revised the project database and included only data that were
considered reasonable based on engineering judgment. Information presented in Figures 1 through 15
shows that the selected projects represented typical Arizona pavement designs, construction practices,
and site conditions.
Extract and Review Distress and IRI Data for Each Selected Project
Table 8 presents the distress and IRI data obtained from LTPP, ADOT PMS, and ADOT research projects.
The ADOT PMS data were augmented by distress information obtained through a review of ADOT
distress videos archived since the late 1990s and field distress surveys conducted by Applied Research
Associates, Inc. (ARA) staff. Distress and IRI data from the research projects were obtained from surveys
conducted as part of this project. Researchers reviewed all distress and IRI data for accuracy,
26
reasonableness, and consistency. A key consideration was whether the raw data as measured and
reported could be converted into the MEPDG reporting units for each performance indicator.
Data Collection and Reporting
Since MEPDG performance indicators’ measuring units were consistent with LTPP distress and IRI data
collection protocols, by default all LTPP projects reported distress and IRI data that could easily be
converted into MEPDG‐compatible units.
Table 8. Distress and IRI Data Sources for Local Calibration and Validation
Pavement Type
Distress/IRI Distress and IRI Data Source
LTPP Projects** ADOT PMS Projects
ADOT Research Projects
New HMA
and HMA
overlaid HMA
Alligator cracking* MON_DIS_AC_REV Distress video
and field survey ADOT/WRI
Transverse
cracking* MON_DIS_AC_REV
Distress video
and field survey ADOT/WRI
Rutting dbo_SodaMaster ADOT/WRI
IRI MON_PROFILE_MASTER dbo_SodaMaster ADOT/WRI
New JPCP
Transverse cracking MON_DIS_JPCC_REV Distress video
and field survey ADOT SPR 264
Transverse joint
faulting
MON_DIS_JPCC_FAULT_
SECT NA ADOT SPR 264
IRI MON_PROFILE_MASTER dbo_SodaMaster ADOT SPR 264
Composite
(ARFC
overlaid JPCP)
Alligator cracking MON_DIS_AC_REV Distress video
and field survey N/A
Transverse
cracking* MON_DIS_AC_REV
Distress video
and field survey N/A
Rutting MON_T_PROF_INDEX_
SECTION dbo_SodaMaster N/A
IRI MON_PROFILE_MASTER dbo_SodaMaster N/A
N/A = not available.
*Includes reflection cracking.
**Data tables listed were obtained from the LTPP database, version 25 (January 2011).
27
Results of the ADOT measured distress and IRI data review follow:
HMA‐surfaced pavements:
o Alligator cracking. Data were not available in the ADOT PMS dbo_SodaMaster data table in
a format that could be used for analysis. Alligator cracking distress data were obtained by
viewing archived ADOT PMS distress survey videos and conducting windshield surveys.
o Transverse cracking. Data were not available in the ADOT PMS dbo_SodaMaster data table
in a format that could be used for analysis. Transverse cracking data were also obtained by
viewing archived ADOT PMS distress survey videos and conducting windshield surveys.
o Rutting. Although data were available in the ADOT dbo_SodaMaster data table,
measurements were made using three‐point laser equipment, which is different from LTPP
wire or straight‐edge measurements. After comparing the ADOT and LTPP rutting
measurements to determine if they were compatible, the researchers found that LTPP
rutting measurements were approximately 28 percent higher than ADOT rutting
measurements (Figure 16). ADOT rutting measurements were corrected to make them more
comparable to LTPP.
o IRI. ADOT reported smoothness in terms of IRI.
Much of the recent years’ ride measurement has been in terms of IRI, as defined by LTPP, so these
values were used directly.
y = 1.28xR² = 0.2504
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1 1.2
LTPP rut dep
th, in
ADOT rut depth, in
Figure 16. Relationship Between ADOT and LTPP Rutting Measurements
28
JPCP:
o Transverse cracking. Data were not available in the ADOT PMS dbo_SodaMaster data table
in a format that could be used for analysis. Transverse cracking distress data were obtained
by viewing archived ADOT PMS distress survey videos and conducting windshield surveys.
o Transverse joint faulting. Data were not available in the ADOT PMS dbo_SodaMaster data
table in a format that could be used for analysis.
o IRI. ADOT reported smoothness in terms of IRI.
Composite (ARFC overlaid JPCP) pavements:
o Alligator cracking. None was identified in windshield surveys or from ADOT videos.
o Transverse cracking. Primarily Data suggest that transverse cracking occurred primarily as
transverse joint reflection cracking. No transverse midpanel cracks were identified.
o Rutting. Like HMA pavements, LTPP rutting measurements in composite pavements were
approximately 28 percent higher than ADOT rutting measurements. ADOT rutting
measurements were corrected to make them more comparable to LTPP.
o IRI. ADOT reported smoothness in terms of IRI.
Comparison of Performance Indicator Magnitudes to the Design Threshold (Trigger) Values
The researchers compared the magnitudes of time‐series distress and IRI data from the identified LTPP,
ADOT PMS, and Arizona research projects with design threshold values for each distress type and IRI to
determine whether distress and IRI from the identified projects were typical for Arizona. Results of the
comparison follow:
Average maximum distress and IRI for the projects selected ranged from 10 to 77 percent, which
suggests that for both HMA‐ and JPC‐surfaced pavements, most pavements were in relatively
better condition than the design threshold values. This is typical for in‐service pavements, as
moderately to badly deteriorated pavements are rehabilitated as soon as possible.
Some of the projects selected, however, exhibited distress and IRI values that were significantly
higher than the threshold values, as shown in Figures 17 through 23.
For all the distress types and IRI, the maximum distress and IRI values were greater than the
design threshold values. Thus, the range of distress and IRI values from the identified projects
covered ADOT design threshold values.
Table 9 and Figures 17 through 23 summarize the results of the comparison.
29
Figure 17. Measured Alligator Cracking in LTPP and PMS Sections of
New HMA and HMA Overlaid HMA Projects
Figure 18. Measured HMA Transverse Cracking in LTPP and PMS Sections of
New HMA and HMA Overlaid HMA Projects
30
Figure 19. Measured Rutting in LTPP and PMS Sections of
New HMA and HMA Overlaid HMA Projects
Figure 20. Measured AC IRI in LTPP and PMS Sections of
New HMA and HMA Overlaid HMA Projects
31
Figure 21. Measured JPCP Transverse Cracking in LTPP and
PMS Sections of New JPCP Projects
Figure 22. Measured JPCP Transverse Joint Faulting in LTPP and
PMS Sections of New JPCP Projects
32
Figure 23. Measured JPCP IRI of LTPP and PMS Sections of New JPCP Projects
Table 9. Comparison of Distress and IRI Values with Design Criteria or Threshold Values
Pavement Type
Distress or Performance Indicator
Design Criteria
Maximum Value Statistics for All Projects
Average Value
Percentage of Design Criteria (%)
HMA and HMA overlaid
HMA
Alligator cracking (percent lane
area) 10 6.7 67.0
Transverse thermal cracking
(ft/mi) 1000 775.7 77.6
Rutting (inch) 0.4 0.2 50.0 IRI (inch/mi) 169 65.3 38.6
JPCP
Transverse cracking (percent slabs cracked)
10 3.7 37.0
Transverse joint faulting (inch)
0.15 0.015 10.0
IRI (inch/mi) 169 94.212 55.6
Evaluating Distress/IRI Data to Identify Anomalies and Outliers
Before determining MEPDG inputs, the research team visually inspected time series plots of distress and
IRI data for each LTPP and ADOT PMS project to determine if observed trends in distress and IRI
progression were reasonable and to identify potential anomalies (e.g., significant decrease in distress
33
and IRI magnitude, indicating an occurrence of significant rehabilitation or maintenance event) and
outliers. Figures 24 through 27 are examples of these plots.
IRI,
in/m
i
0
50
100
150
200
250
300
Age, years
0 5 10 15 20 25 30 35 40
SectID=PMS_03-31
PLOTADOT PMS LINEAR NONLINEAR
Figure 24. Progression of Measured IRI with Age in ADOT PMS 03‐31
Mean R
ut D
epth
, in
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Age, years
0 5 10 15 20 25 30 35 40 45 50
SHRPID=4_0509
Figure 25. Progression of Measured Rutting with Age in ADOT LTPP 0509
34
Tra
nsve
rse
Cra
ckin
g, ft/m
i
0
1000
2000
3000
4000
5000
6000
7000
8000
Age, years
0 5 10 15 20 25 30 35 40 45 50
SHRPID=4_0113
Figure 26. Progression of Measured HMA Transverse Cracking with Age in ADOT LTPP 0113
Pe
rce
nt
Sla
bs
Cra
cke
d
0102030405060708090
100
Age, years
0 5 10 15 20 25 30 35 40 45 50
SHRPID=4_0217
Figure 27. Progression of Measured JPCP Transverse Cracking with Age in ADOT LTPP 0217
35
The results of this exercise follow:
Projects exhibiting unreasonable trends in distress and IRI progression were removed and not
used in the analysis. Each distress type and IRI were treated separately, and thus removing a
project from the HMA rutting database, for example, does not imply that it was also removed
from the HMA transverse cracking database.
Individual distress and IRI data (from several years of data for a given project) identified as
outliers and erroneous were removed. Examples include zero measurements that could
represent nonentry values, and significantly high or low distress and IRI values that were
considered unreasonable and not in agreement with observed trends from previous and
subsequent years.
Individual distress and IRI data points measured after the performance of a significant
maintenance or rehabilitation event that significantly altered the pavement design were
removed. Distress and IRI data prior to that point were retained.
Reasonable variability from year to year in distress and IRI measurements was considered
acceptable. This variability is caused by many factors and ultimately contributes to the SEEs used
in defining reliability in predicted distress and IRI.
While initial IRI values (defined as IRI measured within a year of construction or rehabilitation) were
available for some of the projects, other initial IRI values were estimated by extrapolating historical
measurements of IRI back to the construction or overlay completion date.
Final Projects Selection
Following data assembly, review, and cleanup, the researchers selected projects with adequate detailed
information for model verification and local calibration (Table 10). The populated sampling templates
for the new and rehabilitated HMA and JPCP pavements are presented in Tables 11 and 12, respectively.
The number of selected flexible (new HMA and HMA overlaid HMA) pavements and new JPCP projects
exceeded the 18 flexible and 21 JPCP projects required. (See Table 6.) The maps in Figures 28 through 30
show the locations of the identified projects; Table 13 presents a detailed description of these projects.
36
Table 10. Breakdown of Selected Projects for Validation and Local Calibration
Pavement Type LTPP Projects PMS/Research
Projects Total
New HMA pavement 42 16 58*
AC overlay over existing AC pavement 34 8 42*
New JPCP (bare and CPR) 26 22 48*
CRCP (new) 1 1 2
New composite (AC overlaid JPCP) 1 15 16*
AC overlay (old, intact JPCP and
fractured JPCP) 15 0 15*
*Conforms to ADOT pavement design and construction practices.
Table 11. Populated Sampling Template for New and Rehabilitated HMA‐Surfaced Pavements
HMA
Thickness
(inches)
Granular Base
Thickness*
(inches)
Subgrade Type
Coarse‐Grained Soils (AASHTO
Class A‐1 through A‐3)
Fine‐Grained Soils (AASHTO
Class A‐4 through A‐7)
<8
<6
161, 902, 903,
A901, A902, A903,
PMS_98‐115
>6
113, 114, 119, 120, 121, 501,
502, 509, 559, 560,
1007_1, 1021_1, 1034_1,
1034_5, 1036_1, 1037_1,
6053_1, 6055_1, 6055_3,
6060_1,
PMS_03‐07, PMS_03‐15,
PMS_03‐52, PMS_03‐59,
PMS_03‐71
AZ1‐WRI, AZ2‐WRI,
AZ3‐WRI, AZ4‐WRI,
505,
PMS_03‐12
>8
<6
115, 116, 117, 118, 123, 124,
162, 260, 261,
1001, 1002_1, 1002_3
PMS_03‐21_1, PMS_03‐21_2,
PMS_03‐31_1, PMS_03‐31_2
>6
122, 503, 504, 506, 507, 508,
1003_1, 1003_3, 1006_1,
1006_2, 1007_4,
1015_1, 1015_2, 1016_1,
1016_3, 1017_1, 1017_3,
1021_5, 1022_1, 1022_3,
1024_1, 1024_7,6054_1,
B901, B902, B903,
B959, B960, B961, B964
6060_5,
PMS_03‐28
1018_1, 1018_4,
PMS_03‐60
*All projects had a granular base. An equivalent, dense‐graded aggregate base was assumed for PATBs.
37
Table 12. Populated Sampling Template for New and Rehabilitated JPCP
PCC Thickness (inches)
Dowel Diameter
Edge Support*
Subgrade and Base Type
Coarse‐Grained Subgrade Soils (AASHTO Class A‐1
through A‐3)
Fine‐Grained Subgrade Soils (AASHTO Class A‐4 through A‐7)
Lean Concrete Base
Granular‐ or Asphalt‐
Treated Base
Lean Concrete Base
Granular‐ or Asphalt‐
Treated Base
<10
No dowels None
ERES_AZ1‐1, ERES_AZ1‐6, ERES_AZ1‐7,
ERES_AZ 360‐01, ERES_AZ 360‐02, ERES_AZ 360‐03
0601, 0602
Yes 262, 263
Doweled None
218, 7614 214, 222 ERES_AZ‐2, ERES_AZ 10‐04, ERES_AZ 10‐05, ERES_AZ 10‐06, ERES_AZ 10‐07
Yes 217 213, 221, 268
>10
No dowels None
ERES_AZ1‐2,ERES_AZ1‐5, PMS_04‐39
Yes 264, 265,7613,
Doweled None 219 160, 215, 223
Yes 220 216, 266, 267 224,
PMS_88‐68 *Tied PCC and or widened lanes.
38
Figure 28. Location of Selected New HMA and HMA Overlaid HMA Pavement Projects
(LTPP = blue; PMS = red) (base maps product of Microsoft Corporation, copyright 2003)
SPS -1 & - 9
WRI
SPS-5
39
Figure 29. Location of Selected New Bare JPCP Projects (base maps product of Microsoft Corporation, copyright 2003)
SPS
40
Figure 30. Location of Selected New ARFC‐Surfaced JPCP Composite Projects
(LTPP = blue; PMS = red) (base map product of Microsoft Corporation, copyright 2003)
SPS-6
41�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration
(GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
Smith et al.
1991 New JPCP over GB
AZ‐10‐01 N/A 1968 Maricopa
Interstate 10 154.05 Ibid. New JPCP over GB AZ‐10‐02 N/A 1968 Maricopa Interstate 10 154.4 Ibid. New JPCP over GB AZ‐10‐03 N/A 1968 Maricopa Interstate 10 153.87 Ibid. New JPCP over LCB AZ‐10‐04 N/A 1986 Maricopa Interstate 10 140.67 Ibid. New JPCP over LCB AZ‐10‐05 N/A 1985 Maricopa Interstate 10 136.68 Ibid. New JPCP over LCB AZ‐10‐06 N/A 1984 Maricopa Interstate 10 130.88 Ibid. New JPCP over LCB AZ‐10‐07 N/A 1984 Maricopa Interstate 10 130.5 Ibid. New JPCP over CTB AZ1‐1 N/A 1972 Maricopa State (U.S.) 360 (60) 172 Ibid. New JPCP over subgrade AZ1‐2 N/A 1975 Maricopa State (U.S.) 360 (60) 174.3 Ibid. New JPCP over subgrade AZ1‐5 N/A 1979 Maricopa State (U.S.) 360 (60) 177.5 Ibid. New JPCP over LCB AZ1‐6 N/A 1981 Maricopa State (U.S.) 360 (60) 179.5 Ibid. New JPCP over LCB AZ1‐7 N/A 1981 Maricopa State (U.S.) 360 (60) 181.3 Ibid. New JPCP over GB AZ‐17‐01 N/A 1961 Maricopa Interstate 17 203.5 Ibid. New JPCP over GB AZ‐17‐03 N/A 1965 Maricopa Interstate 17 211.89 Ibid. New JPCP over GB AZ‐17‐04 N/A 1965 Maricopa Interstate 17 208.2 Ibid. New JPCP over GB AZ‐17‐11 N/A 1965 Maricopa Interstate 17 208.7 Ibid. New JPCP over LCB AZ‐2 N/A 1985 Maricopa Interstate 10 141.19 Ibid. New JPCP over LCB AZ‐360‐01 N/A 1985 Maricopa State (U.S.) 360 (60) 186.3 Ibid. New JPCP over LCB AZ‐360‐02 N/A 1985 Maricopa State (U.S.) 360 (60) 185.3 Ibid. New JPCP over LCB AZ‐360‐03 N/A 1983 Maricopa State (U.S.) 360 (60) 183.3 LTPP New AC over GB 0113 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over GB 0114 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0115 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0116 N/A 1993 Mohave U.S. 93 52.62
42�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration (Continued)
(GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
LTPP New AC over ATB 0117 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0118 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0119 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0120 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0121 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0122 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0123 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over ATB 0124 N/A 1993 Mohave U.S. 93 52.62
LTPP New JPCP over GB 0160 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over GB 0161 N/A 1993 Mohave U.S. 93 52.62
LTPP New AC over GB 0162 N/A 1993 Mohave U.S. 93 52.62
LTPP New composite (AC/RCC) 0163 N/A 1993 Mohave U.S. 93 52.62
LTPP New JPCP over GB 0213 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over GB 0214 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over GB 0215 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over GB 0216 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over LCB 0217 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over LCB 0218 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over LCB 0219 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over LCB 0220 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over PATB 0221 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over PATB 0222 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over PATB 0223 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over PATB 0224 N/A 1993 Maricopa Interstate 10 109
43�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration (Continued)
(GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
LTPP New AC over GB 0260 N/A 1993 Maricopa Interstate 10 109
LTPP New AC over GB 0261 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over GB 0262 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over PATB 0263 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over PATB 0264 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over GB 0265 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over ATB 0266 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over ATB 0267 N/A 1993 Maricopa Interstate 10 109
LTPP New JPCP over ATB 0268 N/A 1993 Maricopa Interstate 10 109
LTPP New AC over GB 0501 N/A 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0502 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0503 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0504 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0505 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0506 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0507 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0508 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0509 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0559 1990 1968 Pinal Interstate 8 159.01
LTPP Mill and overlay AC 0560 1990 1968 Pinal Interstate 8 159.01
LTPP New JPCP 0601 1991 1967 Coconino Interstate 40 202.16
LTPP New JPCP 0602 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of intact PCC 0603 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of intact PCC 0604 1990 1967 Coconino Interstate 40 202.16
44�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration (Continued)
(GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
LTPP CPR 0605 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of intact PCC 0606 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of crack and
seat PCC 0607 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of crack and
seat PCC 0608 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of rubblized
PCC 0659 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of crack and
seat PCC 0660 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of crack and
seat PCC 0661 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of crack and
seat PCC 0662 1990 1967 Coconino Interstate 40 202.16
LTPP
PCC overlay of crack and
seat PCC (2‐inch AC
interlayer)
0663 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of intact PCC 0664 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of crack and
seat PCC 0665 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of rubblized
PCC 0666 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of crack and
seat PCC 0667 1990 1967 Coconino Interstate 40 202.16
45�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration (Continued) (GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
LTPP AC overlay of intact PCC 0668 1990 1967 Coconino Interstate 40 202.16
LTPP AC overlay of rubblized
PCC 0669 1990 1967 Coconino Interstate 40 122.29
LTPP New AC over GB 0902 N/A 1993 Mohave U.S. 93 60.14
LTPP New AC over GB 0903 N/A 1993 Mohave U.S. 93 60.14
LTPP New AC over GB 1001 N/A 1978 Maricopa Interstate 10 123.34
LTPP Mill and overlay AC 1002B 1996 1980 Yavapai Interstate 40 145.37
LTPP New AC over GB 1002A N/A 1980 Yavapai Interstate 40 145.37
LTPP Mill and overlay AC 1003 1993 1975 Maricopa Interstate 10 98.53
LTPP New AC over GB 1003 N/A 1975 Maricopa Interstate 10 98.53
LTPP Mill and overlay AC 1006 1993 1976 Maricopa Interstate 10 110.65
LTPP New AC over GB 1006 N/A 1976 Maricopa Interstate 10 110.65
LTPP Mill and overlay AC 1007 1995 1978 Maricopa Interstate 10 115.43
LTPP New AC over GB 1007 N/A 1978 Maricopa Interstate 10 115.43
LTPP Mill and overlay AC 1015 1998 1979 Santa Cruz Interstate 19 18.33
LTPP New AC over GB 1015 N/A 1979 Santa Cruz Interstate 19 18.33
LTPP Mill and overlay AC 1016 1997 1979 Santa Cruz Interstate 19 24.17
LTPP New AC over GB 1016 N/A 1979 Santa Cruz Interstate 19 24.17
LTPP Mill and overlay AC 1017 1997 1976 Pima Interstate 19 32.98
LTPP New AC over GB 1017 N/A 1976 Pima Interstate 19 32.98
LTPP Mill and overlay AC 1018 1997 1976 Pima Interstate 19 36.2
LTPP New AC over GB 1018 N/A 1976 Pima Interstate 19 36.2
LTPP Mill and overlay AC 1021 1997 1978 Mohave Interstate 40 72.87
LTPP New AC over GB 1021 N/A 1978 Mohave Interstate 40 72.87
LTPP Mill and overlay AC 1022 1996 1977 Mohave Interstate 40 77.69
46�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration (Continued) (GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
LTPP New AC over GB 1022 N/A 1977 Mohave Interstate 40 77.69
LTPP Mill and overlay AC 1024B 1999 1977 Yavapai Interstate 40 106.95
LTPP New AC over GB 1024A N/A 1977 Yavapai Interstate 40 106.95
LTPP Mill and overlay AC 1025B 1999 1978 Yavapai Interstate 40 113.03
LTPP New AC over GB 1025A N/A 1978 Yavapai Interstate 40 113.03
LTPP AC over existing AC 1034 2001 1975 La Paz State 95 145.25
LTPP New AC over GB 1034 N/A 1975 La Paz State 95 145.25
LTPP New AC over GB 1036 N/A 1983 Mohave U.S. 93 27.64
LTPP New AC over GB 1037 N/A 1985 Mohave State 68 1.4
LTPP AC over existing AC 1062 1999 1977 Mohave Interstate 40 92.75
LTPP AC over existing AC 1065 1999 1977 Yavapai Interstate 40 97.72
LTPP AC over existing AC 6053 1968 N/A Pima Interstate 10 292.89
LTPP AC over existing AC 6054 1969 N/A Santa Cruz Interstate 19 52.25
LTPP AC over existing AC 6055 1975 N/A Maricopa State 85 141.84
LTPP Mill and overlay AC 6055 1999 1975 Maricopa State 85 141.84
LTPP AC over existing AC 6060 1967 N/A Santa Cruz Interstate 19 14.88
LTPP Mill and overlay AC 6060 1998 1967 Santa Cruz Interstate 19 14.88
LTPP New CRCP over ATB 7079 N/A 1989 Maricopa Other 101 11.9
LTPP New JPCP over subgrade 7613 N/A 1979 Maricopa State (U.S.) 360 (60) 180
LTPP New JPCP over CTB 7614 N/A 1984 Maricopa Interstate 10 130.5
LTPP New AC over GB A901 N/A 1993 Mohave U.S. 93 60.14
LTPP New AC over GB A902 N/A 1993 Mohave U.S. 93 60.14
LTPP New AC over GB A903 N/A 1993 Mohave U.S. 93 60.14
LTPP Mill and overlay AC B901 1995 1978 Maricopa Interstate 10 122.29
47�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration (Continued) (GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
LTPP Mill and overlay AC B902 1995 1978 Maricopa Interstate 10 122.29
LTPP Mill and overlay AC B903 1995 1978 Maricopa Interstate 10 122.29
LTPP Mill and overlay AC B959 1995 1978 Maricopa Interstate 10 122.29
LTPP Mill and overlay AC B960 1995 1978 Maricopa Interstate 10 122.29
LTPP Mill and overlay AC B961 1995 1978 Maricopa Interstate 10 122.29
LTPP Mill and overlay AC B964 1995 1978 Maricopa Interstate 10 122.29
PMS New composite
(ARFC/JPCP) 95‐22
2004 1997 Maricopa
State 202 13.23
PMS New composite
(ARFC/JPCP) 92‐39
1994 1994 La Paz
Interstate 10 62
PMS New composite
(ARFC/CRCP) 92‐39
1994 1994 La Paz
Interstate 10 70.66
PMS New composite
(ARFC/JPCP) 95‐49
2004 2001 Maricopa
State 101 34.5
PMS New composite
(ARFC/JPCP) 96‐36
2003 1999 Maricopa
State 101 41
PMS New AC over CTB 00‐78 N/A 2001 Gila State 260 266
PMS New composite
(ARFC/JPCP) 04‐68 2007 2007 Navajo Interstate 40 253.2
PMS New AC over AB 03‐02 2007 2007 Coconino U.S. 160 319.7
PMS New AC over AB 03‐02 2007 2007 Coconino U.S. 160 319.7
PMS AC over existing AC 98‐115 1999 N/A Apache U.S. 160 465.3
PMS New composite
(ARFC/JPCP) 97‐68 2003 1999 Maricopa State
101 22.69
PMS New composite
(ARFC/JPCP) 99‐13 2003 2001 Maricopa State
101 24.5
48�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration (Continued) (GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
PMS New composite
(ARFC/JPCP) 98‐159 2003 2001 Maricopa State
101 28
PMS New AC over GB 03‐28 N/A 2005 Graham U.S. 191 97.6
PMS Mill and overlay AC 02‐55 2006 2006 La Paz State 95 143.0
PMS Mill and overlay AC 03‐31 2002 1988 Pima State 77 72.5
PMS Mill and overlay AC 03‐31 2002 1988 Pima State 77 72.95
PMS New composite
(ARFC/JPCP) 02‐63 2005 2005 Maricopa State
202 44.6
PMS New AC over GB 03‐71 N/A 2008 Mohave U.S. 93 156.5
PMS New AC over GB 03‐52 N/A 2005 Mohave U.S. 93 122.96
PMS New composite
(ARFC/JPCP) 02‐79 2004 1997 Maricopa Interstate 10 157.7
PMS New AC over GB 03‐59 N/A 2008 Yuma State 195 9.1
PMS New composite
(ARFC/JPCP) 01‐86 2008 2007 Pima Interstate 10 255
PMS New AC over GB 03‐60 N/A 2006 Navajo Interstate 40 287.6
PMS Mill and overlay AC 03‐12 2006 2006 Yavapai Interstate 17 263
PMS AC over existing AC 03‐15 2006 N/A Mohave U.S. 93 95.1
PMS Mill and overlay AC 03‐21 2005 2005 Apache U.S. 180 416.31
PMS Mill and overlay AC 03‐21 2005 2005 Apache U.S. 180 416.31
PMS New composite
(ARFC/JPCP) 03‐06 2005 2005 Maricopa State
202 31
PMS New AC over GB 03‐07 N/A 2004 Gila U.S. 70 252.14
PMS AC over existing AC 03‐31 2009 2002 Pima State 77 72.95
49�
Table 13. LTPP and ADOT PMS Projects Selected for Validation and Local Calibration (Continued)
(GB = granular base; LCB = lean concrete base; ATB = asphalt‐treated base)
Source Pavement Type ARA ID
Construction
or Rehab
Year
Original
Construc‐
tion Year
County Route Signing Route No. Milepost
PMS New composite
(ARFC/JPCP) 02‐24 N/A 2004 La Paz Interstate 10 1.8
PMS New JPCP over AB 04‐39 N/A 2006 Mohave State 66 56.56
PMS New JPCP over ATB 88‐68 N/A 1993 Pima Interstate 10 260.4
PMS ARFC surfacing of CRCP PMS7079 2005 1989 Maricopa State 101 11.9
WRI New AC over GB AZ1‐1 WRI N/A 2001 Maricopa U.S. 93 145.45 WRI New AC over GB AZ1‐2 WRI N/A 2001 Maricopa U.S. 93 146.47 WRI New AC over GB AZ1‐3 WRI N/A 2001 Maricopa U.S. 93 147.13 WRI New AC over GB AZ1‐4 WRI N/A 2001 Maricopa U.S. 93 147.83
50
Extract and Review Key MEPDG Input Data for Rehabilitated Projects
Specifically, for AC overlaid existing HMA and ARFC over JPCP projects, the distress condition of
the existing pavement before overlay and repair work performed as part of rehabilitation is a
key input for modeling future performance. The following information is required for existing
new HMA pavements and existing JPCP:
Existing HMA pavements:
o Alligator cracking.
o Rutting (in all layers of relevance).
o Repair work done (full‐depth patching and milling depth).
Existing JPCP:
o Transverse cracking.
o Slab replacement.
Before overlay placement, the researchers used this information to characterize pavement
condition. The distress and IRI data were obtained as described previously.
STEP 6: CONDUCT FIELD AND FORENSIC INVESTIGATIONS
Forensic investigations were limited to ADOT PMS projects and were composed of either a
windshield distress survey, FWD testing, or both. As noted, the researchers conducted
windshield distress surveys on many ADOT PMS projects with insufficient distress data to
update condition information or to characterize the condition entirely. The researchers also
completed FWD testing and backcalulation of subgrade elastic/resilient modulus for ADOT PMS
sections with no foundation support data.
STEPS 7 THROUGH 10: ASSESS LOCAL BIAS AND STANDARD ERROR OF THE ESTIMATE FROM
GLOBAL CALIBRATION FACTORS, AND ELIMINATE/REDUCE STANDARD ERROR OF THE
ESTIMATE AND LOCAL BIAS OF DISTRESS PREDICTION MODELS
Several methods were used singly or in combination to verify the MEPDG global models for
Arizona local conditions. Statistical methods were used when there was a good distribution of
measured distress and IRI; when measured distress was mostly zero, a nonstatistical approach
to model validation was used. Model validation consisted of the following steps:
1. Execute the MEPDG for each selected LTPP, ADOT PMS, and ADOT research project (see
Table 13) and predict pavement distresses and IRI over a reasonable time period
(typically 10 to 40 years).
51
2. Extract predicted distress and IRI data from the MEPDG outputs that match the age of
measured distress and IRI.
3. Perform statistical or nonstatistical analysis as appropriate to characterize bias and
prediction capacity.
4. Determine suitability of the MEPDG global models for Arizona local conditions.
Details of the statistical or nonstatistical analysis used to verify model suitability are presented
in the following sections.
Evaluate Model Prediction Capacity (Goodness of Fit)
To assess the goodness of fit of a given MEPDG distress and IRI prediction model, the
researchers determined the coefficient of determination (R2) and SEE using measured Arizona
and MEPDG predicted distress and IRI data. Reasonableness of both diagnostic statistics was
determined by comparing the Arizona diagnostic statistics with those obtained for the MEPDG
global models under NCHRP 1‐40D (see Table 14). Engineering judgment was then used to
determine the reasonableness of Arizona diagnostic statistics.
Models exhibiting a poor R2 (i.e., R2 significantly less than the values presented in Table 14) or
excessive SEE (significantly higher than the values presented in Table 14) were considered
inadequate for Arizona conditions.
Table 14. National Calibration Under NCHRP 1‐40D New HMA Pavement and
New JPCP Model Statistics
Pavement Type Performance
Model
Model Statistics
R2 SEE Number of Data
Points (N)
New HMA
Alligator cracking 0.275 5.01% 405
Transverse thermal
cracking
Level 1*: 0.344
Level 2*: 0.218
Level 3*: 0.057
— —
Rutting 0.58 0.107 inch 334
IRI 0.56 18.9 inch/mi 1926
New JPCP
Transverse slab
cracking 0.85 4.52% 1505
Transverse joint
faulting 0.58 0.033 inch 1239
IRI 0.60 17.1 inch/mi 163
*Level of inputs used for calibration.
52
Evaluate Model Bias
Bias was defined as the consistent under‐ or overprediction of distress and IRI. The researchers
determined bias by performing linear regression using Arizona measured and MEPDG predicted
distress and IRI, and performing the following two hypothesis tests. A significance level, , of 0.05 or 5 percent was assumed for all hypothesis testing.
Hypothesis 1: Paired t‐test. This test determined whether the Arizona measured and
MEPDG predicted distress and IRI represented the same population. The paired t‐test
consisted of the following steps:
o Assume the following null and alternative hypothesis:
a. H0: mean measured distress/IRI = mean predicted distress/IRI.
b. HA: mean measured distress/IRI ≠ mean predicted distress/IRI.
o Compute test p‐value using SAS Institute Inc. (SAS) statistical software.
o Compare computed p‐value to predetermined level of significance for this test.
(Note: A significance level of 5 percent was adopted for this analysis.)
The null hypothesis H0 was rejected if the p‐value was less than 0.05. Rejecting H0 implied
that the Arizona measured and MEPDG predicted distress and IRI were essentially from
different populations at the 5 percent significance level. Belonging to different populations
indicates a potential bias in the MEPDG predicted distress and IRI since the measured
distress and IRI are considered to represent the ground truth for this test analysis.
Hypothesis 2. This paired t‐test (which is only meaningful if hypothesis 1 was accepted)
determined whether a linear regression model (predicted distress/IRI = *measured
distress/IRI) has a slope () of 1.0 at the 5 percent significance level. The test consisted of the following steps:
o Using the results of the linear regression analysis, test the following null and
alternative hypotheses to determine if the linear regression model slope is 1.0:
a. H0: model slope () = 1.0.
b. HA: model slope () ≠ 1.0.
o Compute test p‐value using SAS.
o Compare computed p‐value to predetermined level of significance for this test.
(Note: A significance level of 5 percent was adopted for this analysis.)
The null hypothesis H0 was rejected if p‐value was less than 0.05. Rejecting H0 implied that
the linear model has a slope significantly different from 1.0 at the 5 percent significance
level, which indicates that using the MEPDG global models outside of the range of Arizona
measured distress and IRI will produce biased predictions.
53
The presence of bias did not necessarily imply that the MEPDG global models were defective. It
simply means that there is some bias in predicted distress and IRI values when the MEPDG is
used under Arizona conditions.
The global MEPDG models that were considered biased or had an inadequate prediction
capacity or goodness of fit needed local calibration.
Calibrate MEPDG Models for Arizona Local Conditions
The researchers used both linear and nonlinear regression procedures from the SAS statistical
software to calibrate MEPDG models for local conditions:
1. Split the assembled LTPP and ADOT projects data into two separate databases for
calibration and validation. For this study, a 90/10 percent split was adopted. Thus,
approximately 90 percent of all projects were used to create a local calibration database
while the remaining 10 percent of projects were used to create a validation database.
While projects were selected at random, the researchers ensured the projects
represented a balance in pavement type, design features, material properties, traffic,
construction year, location and climate, and other parameters. In essence, the
10 percent validation database had the same distributions and key statistics such as the
range of AC or PCC thickness and mean values of such key inputs. After the preliminary
or tentative locally calibrated MEPDG models were developed and validated using the
90 percent local calibration and 10 percent validation databases, respectively, the final
locally calibrated MEPDG models were developed using all projects. Tables 15 and 16
present the projects selected for the databases.
2. Perform local calibration:
a. Run the MEPDG for all projects in the 90 percent local calibration database.
b. Assemble critical MEPDG inputs (HMA thickness); MEPDG outputs (predicted
damage, distress, and IRI); and measured distress and IRI required for local
calibration in an optimization database.
c. Determine MEPDG models calibration coefficients that can be modified as part of
local calibration (see Chapter 2).
d. Perform optimization using linear and nonlinear regression techniques to select
local calibration coefficients that maximizes R2, minimizes SEE, and eliminates bias.
54
3. Validate the tentative locally calibrated MEPDG models:
a. Run the MEPDG for all projects in the 10 percent validation database using the local
calibration coefficients previously determined.
b. Assemble critical MEPDG inputs (HMA thickness); MEPDG outputs (predicted
damage, distress, and IRI); and measured distress and IRI required for local
calibration in an optimization database.
c. Determine goodness of fit and presence or absence of bias.
4. Modify the tentative locally calibrated MEPDG models as needed until an adequate
model or solution is obtained (i.e., reasonable goodness of fit and no bias).
5. Develop final locally calibrated MEPDG models using 100 percent of all projects.
Chapter 3 of this report provides a detailed description of Steps 7 through 10.
STEP 11: INTERPRET RESULTS AND DETERMINE ADEQUACY OF ADOT MEPDG LOCALLY
CALIBRATED MODELS
Under this step, the project team developed pavement designs for selected typical ADOT
projects using the final locally calibrated MEPDG models. The new designs were then evaluated
for reasonableness using the following criteria:
Selected in‐service pavements that have performed adequately (i.e., presence of little or
no distress over 15 to 30 years): The new designs are expected to be comparable to the
existing pavement design.
Selected in‐service pavements that have exhibited early failures (i.e., presence of high
levels of distress within 10 to 15 years of a 20‐ to 30‐year design): The new designs are
expected to be an improvement (e.g., thicker HMA and PCC layers) of the existing
pavement designs.
These results are presented in Chapter 3.
55
Table 15. Sampling Template for the 10 Percent Validation/90 Percent Calibration
Databases for New and Rehabilitated HMA‐Surfaced Pavements
HMA Thickness (inches)
Base Thickness (inches)
Projects for Validation and Calibration Databases
Subgrade Type
Coarse (AASHTO Class A‐1 through A‐3)
Fine (AASHTO Class A‐4 through A‐7)
4 to 8
<6
10% validation
90% calibration 161, A901, A902, A903,
PMS_98‐115
>6
10% validation 560, 1007_1, 1021_1, 6055_1 AZ1‐WRI
90% calibration
113, 114, 119, 120, 121, 501,
502, 509, 559, 1034_1,
1034_5, 1036_1, 1037_1,
6053_1, 6055_3, 6060_1,
PMS_03‐07, PMS_03‐15,
PMS_03‐52, PMS_03‐59,
PMS_03‐71
AZ2‐WRI,
AZ3‐WRI, AZ4‐WRI,
505
>8
<6
10% validation 115 PMS_03‐21_1
90% calibration
116, 117,
118,123,124,
162, 260, 261,
1001, 1002_1, 1002_3
PMS_03‐21_2,
PMS_03‐31_1,
PMS_03‐31_2
>6
10% validation 1017_3, B901
90% calibration
122, 503, 504, 506, 507, 508,
1003_1, 1003_3, 1006_1,
1006_2, 1007_4, 1015_1,
1015_2, 1016_1, 1016_3,
1017_1
1021_5, 1022_1, 1022_3,
1024_1, 1024_7, 6054_1,
B902, B903, B959, B960, B961,
B964, 6060_5, PMS_03‐28
1018_1, 1018_4,
PMS_03‐60
Note: Nine of 85 new HMA and HMA overlaid HMA pavements were selected for the 10 percent validation database.
56
Table 16. Simplified Sampling Template for the Validation and Local Calibration of New JPCP
PCC Thickness (inches)
Dowels Edge
Support
Projects for Validation and Calibration Databases
Subgrade Type
Coarse (AASHTO Class A‐1 through A‐3)
Fine (AASHTO Class A‐4 through A‐7)
Base Type
Lean Concrete Base
Granular and
Asphalt‐Treated Base/AC
Lean Concrete Base
Granular and Asphalt‐Treated Base/AC
<10
No dowels
None
10% validation ERES_AZ1‐1
90% calibration
ERES_AZ1‐6, ERES_AZ1‐7,
ERES_AZ 360‐01, ERES_AZ 360‐02, ERES_AZ 360‐03
Tied PCC and or widened lanes
10% validation
90% calibration 262,263
Doweled
None
10% validation ERES_AZ‐2
90% calibration
218, 7614
214,222 ERES_AZ 10‐04, ERES_AZ 10‐05, ERES_AZ 10‐06, ERES_AZ 10‐07
Tied PCC and or widened lanes
10% validation
90% calibration
217 213,221, 268
>10
No dowels
None
10% validation
90% calibration ERES_AZ1‐2,
ERES_AZ1‐5, PMS_04‐39
Tied PCC and or widened lanes
10% validation
90% calibration
264,265, 7613
Doweled
None 10% validation 160
90% calibration 219 215,223
Tied PCC and or widened lanes
10% validation 224
90% calibration
220 216,266, 267
PMS_88‐68
Note: Four of 37 new JPCP were selected for the 10 percent validation database.
57
CHAPTER 3. VERIFICATION, LOCAL RECALIBRATION, AND
VALIDATION OF MEPDG MODELS
Verification of the MEPDG nationally calibrated global models and local calibration and
validation results are presented in this chapter. Table 17 lists the models evaluated as part of
Arizona’s MEPDG implementation. The appendix presents detailed descriptions of these models.
NEW HMA AND HMA OVERLAID HMA ALLIGATOR CRACKING
Verification
To verify the MEPDG global alligator cracking models for Arizona conditions, the researchers ran
the MEPDG with the global coefficients for all projects and evaluated goodness of fit and bias.
Figure 31 plots the cumulative fatigue damage versus alligator cracking for all Arizona HMA
sections. The goodness of fit statistics shown in this figure are very poor (R2 = 8.2%, SEE = 14.3%
lane area), and the model predictions are obviously biased. (Under prediction, the curve
represents the global calibration model.) Thus, local calibration of this very important model
was necessary.
Figure 31. Initial Verification of the HMA Alligator Fatigue Cracking Models
with Global Coefficients Using Arizona Performance Data
R2 = 8.2 %
SEE = 14.3 % lane area
N = 363
58�
Table 17. MEPDG Global Models Evaluated for Arizona Local Conditions
Pavement Type
MEPDG Global Models Evaluated
HMA Alligator
Cracking
HMA Transverse
Cracking
Total R
utting
(3 Submodels)
New HMA and
HMA/H
MA IR
I
JPCP Transverse
Cracking
JPCP Transverse Joint
Faulting
New JPCP IR
I
CRCP Punchouts
CRCP IR
I
HMA/H
MA Reflected
Alligator Cracking
HMA/JPCP Reflected
Tran
sverse Cracking
HMA/JPCP Reflected
IRI
HMA/CRCP Reflected
IRI
New HMA
HMA/HMA
JPCP/HMA
CRCP/HMA
New JPCP
HMA/JPCP
PCC/JPCP
(bonded)
JPCP/JPCP
(unbonded)
New CRCP * *
HMA/CRCP
PCC/CRCP
(bonded)
CRCP/CRCP
(unbonded)
*Limited checks and evaluation.
59
Local Calibration
Local calibration involves investigating the causes of poor goodness of fit and bias in MEPDG
nationally calibrated models, and modifying the calibration coefficients of the HMA fatigue
model and the alligator cracking model (see Chapter 2) as needed based on information derived
from Eq. 1 to improve goodness of fit and reduce or eliminate bias. Specifically, the coefficients
k1, k2, and k3 for the HMA fatigue model and C1, C2, and C4 of the alligator cracking model
were modified during the analysis to improve prediction. In HMA overlays of HMA pavements,
the reflection cracking model was also calibrated during the analysis. This model predicts total
alligator cracking at the surface composed of alligator cracking initiating at the bottom of the
HMA overlay and reflected through from the underlying existing HMA layer.
Investigation of Causes of Poor Goodness of Fit and Bias
There were no obvious causes for poor goodness of fit and bias in the global cracking model.
Local Calibration of Alligator Cracking Models
The researchers computed model local coefficients through optimization using SAS statistical
software for the HMA load applications to cracking model:
3322
11ffff k
HMAk
tfHfHMAf ECCkN (Eq. 1)
They also computed model local coefficients for the alligator cracking fatigue damage model (S‐
shaped curve):
))((
4*22
*11160
1BottomDILogCCCCBottom
e
CFC (Eq. 2)
For HMA overlays, they analyzed the percentage of reflected cracks model from the existing
HMA pavement and optimized coefficients determined:
bdtaceRC
1
100 (Eq. 3)
Where
RC = alligator cracks from existing HMA layer reflected, percent reflected
through overlay (0 to 100)
t = age (in years)
a, b, c, d = model coefficients
60
Table 18 provides the preliminary local calibration coefficients and goodness of fit statistics
developed using 90 percent of projects. The results indicate an adequate goodness of fit; several
of the global models coefficients changed.
The preliminary locally calibrated alligator cracking model was then independently validated
using the 10 percent of projects set aside for validation. The results, presented in Table 19, show
an equally good fit to the measured alligator cracking model. For the validation, however, SEE
was significantly higher.
Table 18. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics
for Alligator Cracking Submodels Using 90 Percent of Projects
Model or Submodel Type
Goodness of Fit Model
Coefficients ADOT Local Calibration (Using 90% of Projects)
Fatigue damage
model
R2 = 55.1%
SEE = 12.6% lane
area
N = 332
k1 1.884
k2 3.9492
k3 1.58
Alligator cracking
model
C1 0.75
C2 4.5
C4 6000
Reflection
cracking model
a 3.5+0.75heff
b ‐0.688584‐3.37302*heff‐
0.915469
c 2.7 if heff < 3‐in
4.0 if heff > 3‐in
d 1
Table 19. Goodness of Fit Statistics for Validating the Preliminary 90 Percent Locally
Calibrated Alligator Cracking Submodels with 10 Percent of Projects
Model or Submodel Type
Goodness of Fit Based on 90% of Selected
Calibration Projects
Goodness of Fit Based on 10% of Selected Validation Projects
Fatigue damage
model R2 = 55.1%
SEE = 12.6% lane area
N = 332
R2 = 64.9%
SEE = 21.2% lane area
N = 30
Alligator cracking
model
Reflection cracking
model
61
Given this validation check, the researchers combined the databases and developed the final
locally calibrated alligator cracking models. The models coefficients did not change and
goodness of fit was adequate, as shown in Table 20. The final locally calibrated alligator cracking
model was evaluated for bias. The results, presented in Figure 32, showed the model has no
significant bias.
Table 20. Goodness of Fit and Bias Test Statistics for Final Locally Calibrated
Alligator Cracking Model (Based on 100 Percent of All Projects)
Model Goodness of Fit and Bias
Statistics Model
Coefficients
ADOT Local Calibration
(Using 100% of Projects)
Change from NCHRP 1‐40D Global Models
Fatigue
damage
model
Goodness of Fit
R2 = 50%
SEE = 14.8% lane area
N = 419
Bias Test
H0: Slope = 1.0 (p‐value =
0.9897)
H0: Predicted and measured
cracking from the same
population (paired t test) (p‐value = 0.0837)
k1 0.007566 Yes
k2 3.9492 No
K3 1.281 Yes
BF1 249.0087 Yes
BF2 1.00 No
BF3 1.2334 Yes
Alligator
cracking
model
C1 1.00 No
C2 4.50 Yes
C4 6000 No
Reflection
cracking
model
a 3.5+0.75heff No
b
‐0.688584‐
3.37302*heff‐
0.915469
No
c 2.55 Yes
d 1.23 Yes
62
Figure 32. Measured and Predicted Alligator Cracking Versus Cumulated Fatigue
Damage for the Locally Calibrated Alligator Cracking Submodels
Figure 33 illustrates the model prediction for a new HMA pavement and Figure 34 shows the
model prediction for an overlaid pavement. The impact of local calibration of the HMA
pavement alligator fatigue cracking model is most significant after removing the bias
(underprediction) of prediction, as can be seen when comparing the global coefficients of Figure
31 and the local coefficients in Figure 32. The goodness of fit increased from 8 percent to
50 percent while the SEE increased slightly—from 14.3 to 14.8. These combined effects will
increase the accuracy and eliminate the bias of the alligator fatigue cracking model. HMA
pavement designs based in part on HMA fatigue cracking in Arizona will be more accurate and
optimum (lower cost) at the selected level of design reliability.
N = 419
R2 = 50%
SEE = 14.8% lane area
63
Figure 33. Predicted and Measured Alligator (Fatigue) Cracking for
Arizona I‐8 HMA (LTPP 4_0501) Pavement
Figure 34. Predicted and Measured Alligator (Fatigue) Cracking for
Arizona I‐8 HMA (LTPP 4_0505) Overlaid Pavement
64
NEW HMA AND HMA OVERLAID HMA RUTTING
Verification
To verify the MEPDG global permanent deformation or rutting models for Arizona conditions,
researchers ran the MEPDG with the global coefficients for all projects and then evaluated
goodness of fit and bias. Figure 35 plots predicted versus measured total rutting for all Arizona
HMA sections. The goodness of fit statistics are very poor (R2 = 4.6 %, SEE = 0.31 inch), and the
model predictions are obviously biased (large overprediction using the global calibration model).
Thus, local calibration of this very important model was necessary.
Local Calibration
To calibrate the model, researchers investigated the causes of poor goodness of fit and bias of
the MEPDG nationally calibrated models and modified the local calibration coefficients of the
fatigue damage and alligator cracking submodels as needed based on information derived from
Eq. 1. Specifically, the coefficients k1, k2, and k3 for the HMA pavement rutting submodels and
k1 (aggregate base) and k1 (subgrade) for the aggregate base and subgrade submodels were
modified to improve total rutting prediction.
Figure 35. Predicted and Total Rutting Using Global Coefficients and
Arizona HMA Pavement Performance Data
R2 = 4.6%
SEE = 0.31 inch
N = 479
65
Investigation of Causes of Poor Goodness of Fit and Bias
Although there were no obvious causes for poor goodness of fit and bias in the global model
rutting predictions in Arizona, the following was observed:
For new HMA pavement LTPP GPS projects (constructed mostly in the late 1970s and
early 1980s), the global models underestimated total measured rutting.
For new HMA pavement LTPP SPS projects (constructed mostly in the 1990s), the global
models predicted total rutting reasonably well.
For HMA overlaid HMA pavements, the global models predicted very little rutting in the
existing pavement layers and underestimated predicted total rutting, which consisted
almost entirely of rutting in the HMA overlay.
Goodness of fit was very poor due mostly to the magnitude of variability in year‐to‐year
measurements for individual projects.
Local Calibration of Rutting Submodels
Model local coefficients were determined through optimization using SAS statistical software for
the following model (note that the three main terms in this model are HMA, unbound aggregate
base, and subgrade):
n
r
osubgradevsubgradess
n
r
obasevbaseBB
kkkHMArzrtotalp ehkehkTnk rrrrr
1111)(1)(3322110
(Eq. 4)
All inputs are defined in the appendix.
Table 21 presents the preliminary local calibration coefficients and goodness of fit statistics
developed using 90 percent of projects. These results indicate a large improvement in the
goodness of fit between the global models and the Arizona calibrated models as the R2 is now
18 percent compared to 5 percent, and the SEE is 0.12 inch compared to 0.31 inch. This is still a
fairly poor goodness of fit for the preliminary locally calibrated rutting models. Rutting data only
shows excessive variability for many projects over time. This variability, which is caused by
equipment changes or problems from year to year, contributes to poor goodness of fit and high
standard error.
The preliminary locally calibrated rutting model was then independently validated using the
10 percent of projects set aside for validation. The 10 percent sample randomly selected
66
demonstrated much better fit to the measured rutting model, which indicates that it would be
best to combine all of the data into a 100 percent database for a final calibration (Table 22).
Given this result, the databases were combined to obtain a final locally calibrated rutting model.
For this final calibration, the model coefficients did not change much and goodness of fit was
still poor, as shown in Table 23. The final locally calibrated rutting model was evaluated for bias.
Figure 36 presents the results, which show that the overprediction bias has been removed.
Table 21. Preliminary Local Calibration Coefficients and Goodness of Fit
Statistics for Total Rutting Submodels Using 90 Percent of Projects
Model or Submodel
Type Goodness of Fit
Model Coefficients
ADOT Local Calibration (Using 90% of Projects)
AC rutting
R2 = 18.1%
SEE = 0.123 inch
N = 616
k1 ‐3.491598051 (new HMA)
‐3.190568055 (HMA over HMA)
k2 1.5606
k3 0.235
Base
rutting k1 (base) 0.7905
Subgrade
rutting
k1
(subgrade) 1.116
Table 22. Goodness of Fit Statistics for Validating the Preliminary 90 Percent
Locally Calibrated Total Rutting Submodels with 10 Percent of Projects
Model or Submodel Type
Goodness of Fit Based on 90% of Selected Calibration
Projects
Goodness of Fit Based on 10% of Selected Validation Projects
AC rutting R2 = 18.1%
SEE = 0.123 inch
N = 616
R2 = 49.4%
SEE = 0.107 inch
N = 58
Base rutting
Subgrade rutting
67
Table 23. Goodness of Fit and Bias Test Statistics for Final Locally Calibrated
Rutting Model (Based on 100 Percent of All Projects)
Model or Submodel
Type
Goodness of Fit and Bias Statistics
Model Coefficients
ADOT Local Calibration (Using 100% of Projects)
Change from NCHRP 1‐40D Global Models
AC rutting
Goodness of Fit
R2 = 16.5%
SEE = 0.11 inch
N = 497
Bias Test
H0: Slope = 1.0 (p‐value
= 0.0521)
H0: Predicted and
measured cracking
from the same
population (paired t
test) (p‐value = 0.0568)
K1 ‐3.3541 No
K2 1.5606 No
K3 0.4791 No
BR1 0.69 Yes
BR2 1 No
BR3 1 Yes
Base
rutting
K1 (base) 2.03 No
BS1 (base) 0.14 Yes
Subgrade
rutting
K1 (subgrade) 1.35 No
BS1
(subgrade) 0.37 Yes
Figure 36. Measured and Predicted Total Rutting for New HMA and HMA Overlaid HMA
Pavements for the Locally Calibrated Rutting Submodels (Using 100 Percent of Projects)
N = 497
R2 = 16.5%
SEE = 0.11 inch
68
The poor goodness of fit was due to excessive variability in measured year‐to‐year rutting data
used for the analysis, not because of a major weakness in the models. An example of excessive
variability in rutting measurements over time for one section is shown in Figure 37.
Figure 38 illustrates the model prediction for a typical HMA pavement. The impact of local
calibration of the HMA pavement rutting models is most significant after removing the large
overprediction bias, as can be seen when comparing the global coefficients in Figure 35 and the
local coefficients in Figure 36. The goodness of fit increased from 5 percent to 17 percent. The
SEE reduced significantly, from 0.31 inch to 0.11 inch. HMA pavement designs based in part on
HMA pavement rutting in Arizona will be much more accurate and optimum (lower cost) at the
selected level of design reliability.
Ru
t D
ep
th,
in
0.00.10.20.30.40.50.60.70.80.91.0
Age, months
0 36 72 108 144 180 216 252 288 324 360
SHRPID=4_0162
PLOTDARWin-ME ADOT DARWin-MELTPP WIRE
Figure 37. High Variation of Measured Rutting for LTPP Project 4_0162
Figure 38. Measured and Predicted (Global and Arizona Local Calibrated Models)
Total Rutting for Arizona LTPP Project 4_0113
Dots: measured data
Line: prediction curve
69
TRANSVERSE LOW‐TEMPERATURE CRACKING
The MEPDG’S HMA pavement transverse cracking models are based on low‐temperature
contraction of asphalt binders that leads to tensile stresses and transverse cracks. However,
much of Arizona is not subjected to such low temperatures. Yet the researchers found
transverse cracking at approximately 50‐ to 200‐ft intervals on many LTPP and PMS sections in
nonfrost areas such as the Phoenix and Tucson valleys. Figure 39 shows two LTPP examples—
one in the nonfreeze Phoenix area and the other in the nonfreeze Tucson area—that have
developed extensive transverse cracking over 10 to 15 years. These sections exhibit on the order
of 2000 ft/mi in the time frame, which corresponds roughly to a spacing of 30 ft down the
highway—a significant number of transverse cracks that as they open and deteriorate cause
pavement roughness.
Figure 39. Transverse Cracking in Nonfreeze Areas of Phoenix and Tucson
Sections in cold freeze areas also exhibited transverse cracks with similar spacings. Thus, it
became apparent that more than the traditional low‐temperature transverse cracking
mechanism was occurring in Arizona. Many of the cracks in warm areas were up to 1 inch wide,
causing speculation that significant shrinkage of the HMA mixture, possibly from binder
absorbing into the aggregate, was another mechanism. Of course, the MEPDG cannot handle
this type of mechanism and would therefore underpredict transverse cracking in Arizona’s
nonfreeze areas.
Verification
To verify the MEPDG global transverse cracking model for Arizona HMA conditions, researchers
ran the MEPDG with the global coefficients for the SPS‐1 experimental sections on U.S. 93 and
evaluated goodness of fit and bias. The SPS‐1 site is northwest of Kingman, Arizona, in a
70
moderate temperature zone with 57 air temperature freezing cycles per year and a freezing
index of 249 degree‐days. Figure 40 shows the minimum monthly temperature over the past 10
years at the Kingman airport, which is near the SPS‐1 site. Measured transverse cracking on all
of the SPS‐1 HMA sections after 10 years is shown in Figure 41. The measured transverse
cracking ranges widely from 0 to more than 5000 ft/mi (which is 13‐ft spacing) after 10 years.
Figure 40. Monthly Temperatures Recorded at the SPS‐1 Site Near Kingman, Arizona
Figure 41. Measured Transverse Cracking at the SPS‐1 Site
After 10 Years of Service
71
The MEPDG was then run for all of the HMA sections at the SPS‐1 site using the global
calibration coefficients for Level 3 (e.g., K = 1.5). The results, shown in Figure 42, indicate that
using the global calibration coefficients for Arizona led to an underprediction of transverse
cracking for a range of HMA sections at this site. The model predictions may be biased
(underprediction using the global calibration model with K = 1.5). Thus, local calibration of this
important model was necessary.
Figure 42. Predicted and Measured Transverse Cracking Using
Global Coefficients in SPS‐1 HMA Pavement Sections
Local Calibration
The calibration parameter for Level 3 is K. (See appendix.) This value was varied for the SPS‐1
site until the predicted transverse cracking matched the measured on average. Obtaining a
value of K = 100 provided some match at the upper end of cracking but overpredicted at the
lower end, as shown in Table 24. Figure 43 illustrates predicted versus measured transverse
cracking for all sections over time. These results suggest that while there is obviously little
correlation between measured and predicted transverse cracking at the SPS‐1 site, there is at
least some MEPDG predicted values greater than for a Level 3 coefficient (K = 100). This level of
K appears to be overpredicting at this site.
72
Table 24. Measured and Predicted HMA Transverse Cracking
at the SPS‐1 Site (Calibration K = 100)
LTPP Project ID
Survey Date LTPP Transverse Cracking (ft/mi)
MEPDG Transverse Cracking (ft/mi [kt = 100])
0113 April 12, 2005 3257 1843
0114 Nov. 28, 2005 1705 1654
0118 Nov. 28, 2005 124 1295
0122 Dec. 1, 2005 255 1499
0124 Nov. 28, 2005 695 934
Figure 43. Predicted (Calibration K = 100) Versus Measured
Transverse Cracking in SPS‐1 HMA Pavement Sections
To determine the effect of a lower K (e.g., 50), the researchers ran the MEPDG for a variety of
weather station locations, ranging from low elevation hot desert to very cold high elevation
mountain sites, for a typical HMA pavement section (6‐inch HMA with PG 64‐22 and PG 76‐16,
12‐inch aggregate base, and A‐2‐4 coarse‐grained subgrade). Key findings of this effort,
summarized in Table 25, follow:
Predicted low‐temperature transverse cracking varies from 0 to more than 2000 ft/mi
over 10 years for this wide range of climates. Transverse cracking varies directly with
elevation (as shown in Figure 44).
73
Warm nonfreeze climates of Phoenix and Scottsdale with 0 to1 air freeze‐thaw cycles
per year predict no transverse cracking.
Tucson’s slightly cooler climate, with nine air freeze‐thaw cycles per year, predicts
transverse crack spacing of about 737 ft/mi or 75‐ft spacing. Many local streets in
Tucson have transverse crack spacings in this range.
Colder climates of Safford, Kingman, Nogales, and Bisbee, with 32 to 57 air freeze‐thaw
cycles per year, predict 1831 to 1966 ft/mi or 31‐ to 35‐ft transverse crack spacing.
Very cold climates of Winslow, St. Johns, and Window Rock, with 119 to 173 air freeze‐
thaw cycles per year, all predict transverse crack spacing of 2112 ft/mi or 30‐ft crack
spacing.
Table 25. MEPDG Prediction of HMA Transverse Cracking for
Arizona Climates (Calibration K = 50)
Site Location Elevation
(ft)
Air Freeze‐Thaw
Cycles/Year
Freezing Index (inches)
Predicted Transverse Cracks (ft/mi,
PG 76‐16/ PG 64‐22)
Predicted Transverse Crack Space (ft, PG 76‐16/ PG 64‐22)
Phoenix 1105 0 0 0/4 None/None
Scottsdale 1473 1 1 23/29 2754/2182
Tucson 2549 9 21 843/737 75/86
Safford 3170 32 166 1955/1882 32/34
Kingman 3420 57 249 2028/1831 31/35
Nogales 3908 35 174 1847/1872 34/34
Bisbee 4105 55 324 2055/1966 31/32
Winslow 4886 119 1144 2112/2112 30
St. Johns 5722 115 1181 2112/2112 30
Window Rock 6733 173 2157 2112/2112 30
74
Figure 44. MEPDG Predicted Transverse Cracking at Various
Arizona Locations (Calibration K = 50)
These results show that the MEPDG can predict reasonable results for low‐temperature
cracking, at least for areas with cold temperatures and elevations of 3000 ft and higher. In areas
like the Phoenix valley and Tucson valley, where transverse cracking exists on some HMA
pavements, another mechanism such as HMA shrinkage may well be causing transverse cracks
to develop and widen over time.
At this time, it is not recommended to use the MEPDG predicted transverse cracking as a design
performance criteria, similar to alligator cracking or rutting. A calibration coefficient of K ranging
from 7.5 to 50 appears reasonable. Transverse cracking should be observed and perhaps the PG
grade of the binder and other HMA mix properties modified to minimize the potential
development of transverse cracking. Additional research is needed to provide much stronger
verification of the HMA transverse cracking model in Arizona.
NEW HMA AND HMA OVERLAID HMA SMOOTHNESS (IRI)
Verification
To verify the MEPDG global IRI model for Arizona HMA conditions, the researchers ran the
MEPDG with the global coefficients for all projects and evaluated goodness of fit and bias.
Figure 45 plots the predicted versus measured IRI for all Arizona HMA sections. The goodness of
fit statistics shown in Figure 45 is poor (R2 = 30%, SEE = 18.7 inch/mi), and the model predictions
are obviously biased (large overprediction using the global calibration model for lower IRI and
underprediction for higher IRI). Thus, local calibration of this very important model was
necessary. Details of this analysis are provided in the appendix.
75
Figure 45. Predicted Versus Measured IRI Using Global Coefficients and
Arizona HMA Pavement Performance Data
Local Calibration
To calibrate the model, researchers investigated the causes of poor goodness of fit and bias of
the MEPDG nationally calibrated models and modified the local calibration coefficients of the IRI
model as needed based on information derived from Eq. 1. Specifically, the coefficients of the
distress inputs and site factor (SF) were modified as needed to improve the HMA IRI prediction.
Investigation of Causes of Poor Goodness of Fit and Bias
There were no obvious causes for poor goodness of fit and bias in the global model.
Local Calibration of HMA IRI Model
Model local coefficients were determined through optimization using SAS statistical software for
the HMA IRI model:
RDCTCCFCCSFCIRIIRI Totalo 4321 (Eq. 5)
where all inputs are as already defined.
Table 26 provides the preliminary local calibration coefficients and goodness of fit statistics
developed using 90 percent of projects. The results indicate a large improvement in the
R2 = 30%
SEE = 18.7 inches/mi
N = 675
76
goodness of fit between the global models and the Arizona calibrated models as the R2 is now
81 percent compared to 30 percent, and the SEE is 8 inches/mi compared to 18.7 inches/mi.
This is a very good goodness of fit for the preliminary locally calibrated HMA IRI model. IRI
seems to be more consistently measured year after year as opposed to the high variability of
rutting from year to year.
The preliminary locally calibrated IRI model was then validated using the 10 percent of projects
set aside for validation. The results in Table 27 show a similar fit to the measured rutting model,
validating the 90 percent model as reasonable.
Given this successful validation check, the databases were combined to obtain a final locally
calibrated rutting model. For this final calibration, the model coefficients did not change much
and goodness of fit was still very good, as shown in Table 28. The final locally calibrated HMA
pavement IRI model was evaluated for bias. The results, presented in Figure 46, show that the
overprediction bias has been removed.
Table 26. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics
for HMA IRI Model Using 90 Percent of Projects
Model or Submodel
Type Goodness of Fit
Model Coefficients
ADOT Local Calibration (Using 100% of Projects)
HMA IRI model
R2 = 81.4%
SEE = 8.1 inch/mi
N = 487
C1 0.0202
C2 0.0400
C3 0.008
C4 25.0
Table 27. Goodness of Fit Statistics for Validating the 90 Percent Preliminary
Locally Calibrated HMA IRI Model with 10 Percent of Projects
Model or Submodel Type
Goodness of Fit Based on 90% of Selected
Calibration Projects
Goodness of Fit Based on 10% of Selected Validation
Projects
HMA IRI model
R2 = 81.4%
SEE = 8.1 inch/mi
N = 487
R2 = 81.7%
SEE = 6.1 inch/mi
N = 70
77
Table 28. Goodness of Fit and Bias Test Statistics for Final Locally Calibrated
HMA IRI Model (Based on 100 Percent of All Projects)
Model or Submodel
Type
Goodness of Fit and Bias Statistics
Model Coefficients ADOT Local Calibration
Change from NCHRP 1‐40D Global Models
HMA IRI
model
Goodness of Fit
R2 = 82.2%
SEE = 8.7 inch/mi
N = 559
Bias Test
H0: Slope = 1.0
(p‐value = 0.7705)
H0: Predicted and
measured cracking
from the same
population (paired
t test) (p‐value =
0.1419)
C1 (for rutting) 1.2281 Yes
C2 (for fatigue) 0.1175 Yes
C3 (for transverse
cracking) 0.0080 No
C4 (for SF) 0.0280 Yes
Figure 46. Measured and Predicted HMA IRI for New HMA and
HMA Overlaid HMA Pavements for the Locally Calibrated
IRI Model (Using 100 Percent of Projects)
R2 = 82.2%
SEE = 8.7 inch/mi
N = 559
78
Model IRI predictions for typical HMA pavements are shown in Figures 47 and 48. The impact of
local calibration of the HMA pavement IRI model is most significant in removing the large
overprediction bias, as can be seen when comparing the global coefficients in Figure 45 and the
local coefficients in Figure 46. The goodness of fit increased from 30 percent to 82 percent. The
SEE reduced significantly from 18.7 inches/mi to 9 inches/mi. HMA pavement designs based in
part on HMA pavement IRI in Arizona will be much more accurate and optimum (lower cost) at
the selected level of design reliability.
IRI, in
/mi
0
40
80
120
160
200
240
280
320
Age, months
0 36 72 108 144 180 216 252 288 324 360
SHRPID=4_0260
Figure 47. Measured and Predicted IRI for HMA LTPP Project 4_0260
IRI,
in/m
i
0
40
80
120
160
200
240
280
320
Age, months
0 36 72 108 144 180 216 252 288 324 360
SHRPID=4_0122_calib
Figure 48. Measured and Predicted (Global and Arizona Local
Calibrated Models) Total IRI for Arizona LTPP Project 4_0122
79
NEW AND RECONSTRUCTED JPCP TRANSVERSE CRACKING
Verification
To verify the MEPDG global JPCP transverse cracking for Arizona conditions, the researchers ran
the MEPDG with the global coefficients for all projects and evaluated goodness of fit and bias.
Note that for JPCP projects, the global model coefficients developed under the recently
completed NCHRP project 20‐07 to reflect corrections in concrete CTE values. The corrected CTE
values were also used in the Arizona model calibration.
The analysis first used the Arizona database to establish the goodness of fit and bias in the
MEPDG transverse cracking models. In Figure 49, the predicted versus measured slab cracking
for the global calibration coefficients shows poor goodness of fit along with bias in
underprediction of cracking.
Figure 49. Predicted Versus Measured Percent Slabs Cracked for
Arizona JPCP with Global Calibration Coefficients
R2 = 20%
SEE = 9%
N = 197
80
Local Calibration
To calibrate the model, researchers investigated the causes of poor goodness of fit and bias of
the MEPDG nationally calibrated models and modified the local calibration coefficients of the
transverse fatigue cracking models as needed based on information derived from Eq. 1. Two key
models are involved with the calibration of transverse slab cracking: Eq. 6 estimates the fatigue
life (N) of PCC when subjected to repeated stress for a given flexural strength. Calibration
factors C1 and C2 could be modified but since this is based on substantial field testing data,
changing these coefficients is not recommended.
(Eq. 6)
The most likely model with appropriate coefficients is the S‐shaped curve giving the relationship
between field measured cracking and accumulated fatigue damage at the top and bottom of the
JPCP slabs. Parameters C4 and C5 are the prime candidates to adjust to remove bias and
improve goodness of fit with field data.
(Eq. 7)
Investigation of Causes of Poor Goodness of Fit and Bias
JPCPs with asphalt‐treated or aggregate bases were found to predict transverse cracking
reasonably well (Figures 50 and 51, respectively). There is little cracking measured and the
predictions show a similar amount. The main cause of poor goodness of fit and bias in the global
model was poor prediction of JPCPs constructed over lean concrete bases for a specific section
(Figure 52). The measured fatigue transverse cracking develops far earlier than the MEPDG
predicted models using global coefficients.
2
,,,,,1,,,,,log
C
nmlkji
inmlkji
MRCN
541
1C
FDICCRK
81
Figure 50. Arizona JPCP with HMA Base Course Measured and
Predicted Percent Slabs Transversely Cracked
Figure 51. Arizona JPCP with Unbound Aggregate Base Course Measured
and Predicted Percent Slabs Transversely Cracked
82
Figure 52. Measured and Predicted JPCP Transverse Cracking Using
MEPDG Global Models for JPCP Over Lean Concrete Base
(The slab/base friction was released at 11 years according to DARWin‐ME guidelines, but
Arizona PCC/ lean concrete base construction indicates friction is lost almost immediately after
opening to traffic, which moves the predicted curve back through the data points.)
A close examination of the JPCPs constructed over lean concrete bases transverse cracking
prediction indicated that the global‐derived typical loss of bonding between the PCC and
underlying cement‐treated base/lean concrete base age of 10.6 years was unsuitable for
Arizona. The lean concrete base construction of these sections from SPS‐2 required a smoothed
surface sprayed with wax curing compound and then another spraying just before paving. Thus,
the slab and lean concrete base lost bond and friction very quickly, and the rapid fatigue
damage and transverse cracking process began. Based on the actual trends observed, a loss of
bond age of 0 years was adopted assuming that the PCC and underlying cement‐treated
base/lean concrete base never really bonded.
Local Calibration of Transverse Cracking Model
The Arizona JPCP sections with lean concrete base or cement‐treated bases were modified for
loss of friction age of 0 and all sections rerun. The other base types were all set at full friction
over their service life as before.
Model local coefficients were determined through optimization using SAS statistical software for
the JPCP faulting model. The results showed a very significant improved model goodness of fit
and no bias in the predictions for JPCP with any type of base course, as summarized in Table 29
83
and plotted in Figure 53. Figure 54 shows the computed cumulated PCC fatigue damage and
measured fatigue transverse cracking (percent slabs). As fatigue damage increases measured
cracking doesn’t increase until it reaches a very critical point where fatigue cracking rapidly
increases. These results are similar to the HMA fatigue cracking curve.
Table 29. Goodness of Fit and Bias Test Statistics for Globally Calibrated JPCP
Transverse Cracking Model (Based on 100 Percent of All Projects)
Model Types Goodness of Fit and Bias
Statistics Model
Coefficients
NCHRP 20‐07 Global Model (Using 100% of Projects)
PCC fatigue
allowable N
model
Goodness of Fit
R2 = 72.8%
SEE = 7.25%
N = 198
Bias Test
H0: Slope = 1.0 (p‐value =
0.8840)
H0: Predicted and measured
cracking from the same
population (paired t test)
(p‐value = 0.0504)
C1 2
C2 1.22
JPCP transverse
cracking model
C4 0.19
C5 ‐2.067
Figure 53. Predicted Versus Measured Percent Slabs Transverse Cracked
for Arizona JPCP with All Types of Base Courses
R2 = 72.8%
SEE = 7.25%
N = 198
84
Figure 54. Measured and Predicted JPCP Transverse Cracking Versus Cumulative Fatigue
Damage for Local Arizona Calibration Coefficients (Using 100 Percent of Projects)
Transverse fatigue cracking model predictions are shown in Figure 55 (lean concrete base),
Figure 56 (untreated aggregate base), and Figure 57 (HMA base). All show good fit to the
MEPDG calibrated models.
The impact of local calibration of the JPCP cracking model is most significant in removing the
large bias that existed with the lean concrete base. The goodness of fit increased from
20 percent to 73 percent. The SEE reduced significantly from 9 percent to 7 percent. JPCP
designs based in part on transverse cracking in Arizona will be much more accurate and
optimum (lower cost) at the selected level of design reliability.
R2 = 72.8%
SEE = 7.25%
N = 198
85
Figure 55. Predicted (Using Local Calibration Factors and Input Recommendations) and
Measured Transverse Cracking for JPCP 4_0217 (SPS‐2) with Lean Concrete Base
(Slab/base friction was released at time 0 opening to traffic for this prediction curve.)
Figure 56. Predicted (Using Local Calibration Factors and Input Recommendations) and
Measured Transverse Cracking for JPCP 4_0214 (SPS‐2) with Untreated Aggregate Base
(Slab/aggregate base friction is full over entire period.)
86
Figure 57. Predicted (Using Local Calibration Factors and Input Recommendations) and
Measured Transverse Cracking for JPCP 4_0224 (SPS‐2) with HMA Base
(Slab/HMA base friction is full over entire period.)
NEW JPCP TRANSVERSE JOINT FAULTING
Verification
To verify the MEPDG global JPCP transverse joint faulting for Arizona conditions, the researchers
ran the MEPDG with the global transverse joint faulting model for all projects and then
evaluated goodness of fit and bias. Results in Figure 58 show that goodness of fit was fair along
with the SEE but that the model predictions were biased (overprediction).
87
Figure 58. Predicted Versus Measured Joint Faulting for
JPCP with Global Calibration Coefficients
Local Calibration
To calibrate the model, researchers investigated the causes of poor goodness of fit and bias of
the MEPDG nationally calibrated models and modified the local calibration coefficients of the
joint faulting models as needed based on information derived from Eq. 1.
Investigation of Causes of Poor Goodness of Fit and Bias
There were no obvious causes for poor goodness of fit and bias in the global faulting model.
Local Calibration of Transverse Joint Faulting Model
Model local coefficients were determined through optimization using SAS statistical software for
the JPCP faulting model:
m
iim FaultFault
1
(Eq. 8)
iiii DEFaultFAULTMAXCFault *)(* 21134 (Eq. 9)
65
170 C *C CEROD
m
jji LogDEFAULTMAXFAULTMAX
(Eq. 10)
6
)*
(*)0.5*1(* *C 2005curling120
C
s
EROD
P
WetDaysPLogCLogFAULTMAX
(Eq. 11)
where all inputs are as defined in the appendix.
R2 = 45%
SEE = 0.026 inch
N = 199
88
Table 30 provides the preliminary local calibration coefficients and goodness of fit statistics
developed using 90 percent of projects. The results indicate a reasonably good goodness of fit;
most calibration coefficients were changed. SEE was 0.0284 inches and considered adequate.
The preliminary locally calibrated JPCP faulting model was then independently validated using
the 10 percent of projects set aside for validation. Results in Table 31 show a very good
correlation coefficient between the measured and predicted JPCP faulting, and SEE was also
lower.
With the successful validation of the preliminary Arizona calibrated JPCP faulting model, the
researchers developed a final locally calibrated total rutting prediction model using the entire
database of projects. Results in Table 32 show that the model’s coefficients did not change, and
goodness of fit remained essentially the same. The final Arizona calibrated JPCP faulting model
was evaluated for bias and the results indicated it has no significant bias.
Figure 59 shows the direct comparison of predicted and measured joint faulting for 100 percent
of the data.
Table 30. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics for
Transverse Joint Faulting Model Using 90 Percent of Projects
Model and Submodel Types
Goodness of Fit Model
Coefficients ADOT Local Calibration (Using 90% of Projects)
All faulting submodels
Goodness of Fit R2 = 53.8% SEE = 0.0284 inch N = 177
C1 0.0355 C2 0.1147 C3 0.00436 C4 1.1E‐07 C5 20000 C6 2.0389 C7 0.1890 C8 400
Table 31. Goodness of Fit Statistics for Validating the 90 Percent Preliminary Locally Calibrated
Transverse Joint Faulting Model with 10 Percent of the Data
Model and Submodel Types
Goodness of Fit Based on 90% of Selected Calibration Projects
Goodness of Fit Based on 10% of Selected Validation Projects
All faulting
submodels
R2 = 53.8%
SEE = 0.0284 inch
N = 177
R2 = 73.4%
SEE = 0.002 inch
N = 19
89
Table 32. Goodness of Fit and Bias Test Statistics for Final Locally Calibrated Transverse Joint
Faulting Model (Based on 100 Percent of All Projects)
Model and Submodel Types
Goodness of Fit and Bias Statistics Model
Coefficients
ADOT Local Calibration (Using 100% of Projects)
All faulting submodels
Goodness of Fit R2 = 51.6% SEE = 0.0225 inch N = 197 Bias Test H0: Slope = 1.0 (p‐value = 0.0632) H0: Predicted and measured cracking from the same population (paired t test, difference = 0, p‐value = 0.0221; paired t test, difference = 0.001, p‐value = 0.1045)
C1 0.0355 C2 0.1147 C3 0.00436 C4 1.1E‐07 C5 20000 C6 2.0389 C7 0.1890
C8 400
Figure 59. Measured and Predicted Transverse Joint Faulting for Locally
Calibrated Faulting Submodels (Using 100 Percent of Projects)
Figure 60 shows the transverse joint faulting model prediction for a JPCP located on Interstate
10 (I‐10) west of Phoenix under very heavy traffic. The prediction shows good fit to the MEPDG
Arizona calibrated model for transverse joint faulting.
R2 = 51.6%
SEE = 0.0225 inch
N = 197
90
The greatest impact of the Arizona calibration of the JPCP joint faulting model is in removing the
overprediction bias that existed. The goodness of fit increased from 45 to 52 percent, and the
SEE remained about the same. JPCP designs based in part on transverse joint faulting in Arizona
will be more accurate and optimum (lower cost) at the selected level of design reliability.
Figure 60. Measured and Predicted Joint Faulting Using the
Arizona Calibrated Model for JPCP Section 4_0265 on I‐10
NEW JPCP SMOOTHNESS (IRI)
Verification
To verify the MEPDG global JPCP IRI for Arizona conditions, researchers ran the MEPDG with the
global JPCP IRI for all projects and then evaluated goodness of fit and bias. Figure 61 plots
predicted versus measured IRI using the Arizona database; full details of the verification effort
are presented in the appendix. These results indicate that goodness of fit was poor and the
model predictions were biased (overprediction for higher IRI). Thus, local calibration with
Arizona data was required.
91
Figure 61. Predicted JPCP IRI Versus Measured Arizona
JPCP with Global Calibration Coefficients
Local Calibration
To calibrate the model, researchers investigated the causes of poor goodness of fit and bias of
the MEPDG nationally calibrated models and modified the local calibration coefficients of the
JPCP IRI submodels as needed based on information derived from Eq. 1. Specifically, the distress
inputs and SF coefficients were modified as needed to improve predicted JPCP IRI.
Investigation of Causes of Poor Goodness of Fit and Bias
There were no obvious causes for poor goodness of fit and bias in the global model.
Local Calibration of JPCP IRI Model
Arizona‐specific model coefficients were determined through optimization using SAS statistical
software for the JPCP IRI model:
IRI = IRII + J1*CRK +J2*SPALL + J3*TFAULT + J4*SF (Eq. 12)
where all inputs are as already defined.
The results presented in Table 33 indicate a reasonably good goodness of fit for the preliminary
locally calibrated JPCP IRI model developed using 90 percent of the projects. Note that all of the
global models coefficients were changed. The SEE of 19 inches/mi was lower than the
25 inches/mi of the global model.
R2 = 35%
SEE = 24.6 inches/mile
N = 279
92
This model was then independently validated using the 10 percent of projects set aside for
validation. The results, presented in Table 34, show a very good correlation coefficient between
the measured and predicted JPCP IRI data and much lower SEE, which indicates that the model
is predicting the 10 percent independent Arizona data very well.
Table 33. Preliminary Local Calibration Coefficients and Goodness of Fit Statistics
for JPCP IRI Model Using 90 Percent of Selected Calibration Projects
Model and Submodel Types
Goodness of Fit Model
Coefficients ADOT Local Calibration (Using
100% of Projects)
JPCP IRI model
R2 = 57.6%
SEE = 19.1 inches/mi
N = 282
J1 (CRK) 0.60
J2 (SPALL) 3.48
J3 (FLT) 1.22
J4 (SF) 45.20
Table 34. Goodness of Fit Statistics for Validating the 90 Percent Preliminary
Locally Calibrated JPCP IRI Model with the 10 Percent Model
Model and Submodel Types
Goodness of Fit Based on 90% of Selected Calibration Projects
Goodness of Fit Based on 10% of Selected Validation Projects
JPCP IRI model
R2 = 57.6%
SEE = 19.1 inches/mi
N = 282
R2 = 70.6%
SEE = 9.6 inches/mi
N = 28
With the successful validation, the researchers developed a final Arizona‐calibrated JPCP IRI
prediction model using the entire database of projects. For this final calibration, the models
coefficients did not change and goodness of fit remained essentially the same. The final Arizona‐
calibrated JPCP IRI model was evaluated for bias. The results, presented in Table 35, also show
no significant bias.
Predicted versus measured JPCP IRI is shown in Figure 62 for 100 percent of the database.
93
Table 35. Goodness of Fit and Bias Test Statistics for Final Arizona
Calibrated JPCP IRI Model (Based on 100 Percent of All Projects)
Model and Submodel Types
Goodness of Fit and Bias Statistics
Model Coefficients
ADOT Local Calibration (Using 100% of Projects)
JPCP IRI
model
Goodness of Fit
R2 = 80.9%
SEE = 9.85 inches/mi
N = 279
Bias Test
H0: Slope = 1.0 (p‐value =
0.6458)
H0: Predicted and measured
cracking from the same
population (paired t test)
(p‐value = 0.2760)
J1 (CRK) 0.60
J2 (SPALL) 3.48
J3 (FLT) 1.22
J4 (SF) 45.20
Figure 62. Measured and Predicted JPCP IRI (Using 100 Percent of Projects)
Figures 63 and 64 illustrate the JPCP IRI model predictions for various Arizona JPCP sections over
time, showing a good fit to the MEPDG Arizona‐calibrated model for IRI of JPCP pavements.
R2 = 81%
SEE = 9.8 inch/mi
N = 279
94
The Arizona calibration affected the JPCP IRI model by removing the bias and improving the
goodness of fit statistics. Goodness of fit increased from 35 to 81 percent, and the SEE
decreased from 25 inches/mi to 10 inches/mi for the global model. JPCP designs based in part
on IRI in Arizona will be more accurate and optimum (lower cost) at the selected level of design
reliability.
Figure 63. Predicted and Measured JPCP IRI for Arizona Section 04_0262
Figure 64. Predicted and Measured JPCP IRI for Arizona Section 04_0267
95
ARFC/JPCP COMPOSITE PAVEMENT MODELS
The MEPDG uses new HMA and new JPCP models for composite pavement design. This section
addresses verification and validation of the various HMA and JPCP models applied to thin ARFC
over JPCP.
Definition of Composite Pavement
Arizona has constructed many new composite pavement sections with a thin ARFC on a JPCP.
The ARFC is currently 1 inch thick (unlike some older courses that are 0.75 inch thick), ground
tire rubber modified asphalt binder, and gap graded with about 20 percent air voids. Many of
these ARFC have been placed within a few months after the concrete slab construction, but
some were placed after a few years of construction. The ARFC is placed primarily to reduce
pavement and tire noise levels, but it also provides reduced splash and spray during rainstorms,
high friction, and a very smooth surface. The ARFC can also be removed and replaced rapidly
when needed.
The JPCP typically ranges from 11 to 14 inches thick and was designed without consideration of
the ARFC surface. JPCP is doweled with perpendicular joints on the heavier trafficked I‐10 and
Interstate 40 (I‐40) routes and nondoweled on the lower truck trafficked urban Loop 101 and
202 freeways in the Phoenix area, and random skewed joints (at 13‐, 15‐, and 17‐ft intervals).
The base course has been both unbound granular material and HMA mixture.
The oldest composite section included in the Arizona MEPDG calibration database was
constructed on I‐10 west of Phoenix in 1994. The section is still in service with the original ARFC
intact and zero transverse fatigue cracks performing very well. This good performance exists on
several other sections along I‐10 from the California state line to Tucson. The youngest was built
on Loop 202 in 2007 and is also performing very well. The entire Loop 101 and Loop 202 in the
Phoenix metropolitan area were also constructed as bare JPCP, and ARFC surfaces were placed
one to seven years later.
Design of Composite Pavements Using MEPDG
The MEPDG software does not directly consider a composite pavement for new design. It must
be designed as an HMA overlay over an existing JPCP or CRCP. The time between placement of
the JPCP or CRCP can be as short as one month or as late as several years. A future version of
MEPDG will likely be modified to directly consider composite pavement as an original new
design alternative. For now, the user must select HMA overlay of JPCP or CRCP to design a new
composite pavement and specify the dates for construction of the concrete slab and HMA
overlay as close as one month apart.
96
Validation of Distress/IRI Models for Composite Pavements
The key distress models for ARFC over JPCP that are output by DARWin ME include:
ARFC rutting.
Transverse fatigue cracking of the JPCP slabs, similar to bare JPCP.
Smoothness through IRI.
Reflection cracking of the transverse joints.
Some of these distresses and IRI are measured in the LTPP program and in the ADOT PMS,
including rutting and IRI. Transverse slab fatigue cracking and transverse joint reflection cracking
were measured in the field for all of the composite pavement sections.
Because the composite pavements built in Arizona have relatively thin asphaltic surface layers,
the researchers expected that the thin (0.75 to 1 inch) ARFC layer would not affect performance
dramatically, at least the fatigue life of the JPCP. However, the ARFC does alter the temperature
and moisture gradients in the slab, making them less significant on fatigue life and transverse
cracking. Thus, the verification effort focused on validating the development of transverse
fatigue cracks that reflect through the ARFC surface, and verifying the rutting in ARFC, IRI, and
reflection joint cracks.
Rutting of the ARFC
The thin ARFC layer did not show much rutting in the field, even under very heavy truck traffic
and hot temperatures over a concrete slab. The DARWin ME inputs for this layer include the
following for Levels 2 and 3:
Thickness of the layer. The layers ranged from 0.5 to 1 inch. The MEPDG minimum layer
thickness is 1 inch and thus this was used for all cases.
Gradation of the aggregate. A gap gradation was used that approximately corresponded
to the following: ‐3/4‐inch = 100 percent; ‐3/8 = 99 percent; ‐#4 = 37.7 percent; ‐#200 =
1.4 percent.
Unit weight. 135 pcf.
Binder grade. The actual binder grade used was PG 64‐16 as is stated in most design
reports. The rubber asphalt binder is crumb rubber from tires introduced into the
asphalt binder in order to enhance its material properties. The researchers used the
recommendation from both early testing and empirical observation to select the proper
asphalt rubber binder grade to enter into the DARWin ME—a PG grading that best
97
matched the binder viscosity‐temperature relationship coefficients (Ai and VTSi) for
asphalt rubber binders. PG 70‐40, which best matched the PG 64‐16 asphalt rubber
binder, was used for all sections based on initial recommendations.
Air voids. Actual voids reportedly are about 20 percent; however, the MEPDG cannot
practically use this high void content. A value of 12 percent was used for all ARFC over
the concrete slab, which provided reasonable predictions of actual rutting.
Binder content (volumetric basis). 11 percent.
Poisson’s ratio. This input was calculated based on mixture properties and temperature.
Seventeen ARFC/JPCP or CRCP composite sections were used in the rutting validation. Measured
rutting was obtained from the ADOT PMS database (SODA) for all of the projects. Table 36
shows the information used to compare measured rutting with DARWin ME predicted rutting.
The researchers conducted the validation analysis by first observing the relationship between
the numbers of heavy trucks and measured rutting of the ARFC, which is plotted in Figure 65.
The graph shows that rutting increases rapidly within the first year but levels out with very little
change over many millions of trucks and years. The magnitude of rutting is also very low. Rutting
is not a significant problem with the 0.75‐inch thick ARFC in Arizona.
Table 36. Arizona Composite Pavement Rutting and IRI Calibration Data
Section IDDate
PCC/ARFCRoute
Start_M
P
End_
MPPavement Type Year
Trucks
only on
ARFC
Millions
IRI
Measured
IRI
Predicted
Rutting
Measured
Rutting
Predicted
01‐86 2002/2003 10 254.43 256.1 0.75‐in ARFC, 14.5 in doweled JPCP, 4‐in ATB 2002 0 0 0
2009/2010 15 77 72 0.12 0.1002‐79 1997/2004 10 157.7 159.2 0.75‐in ARFC, 13‐in doweled JPCP, 3‐in ATB 1997 0 0 0
2009/2010 11 71 59 0.05 0.10
92‐39 1994/1994 10 62 63.48 0.5‐in ARFC, 14‐in doweled JPCP, 4‐in ATB 1994 0 0 0
2009/2010 17 54 67 0.09 0.11
02‐24 2004/2004 10 1.8 11.7 0.75 in ARFC, 13‐in dowled JPCP, AB 2004 0 0 0
2009/2010 9 60 61 0.08 0.1104‐68 2007/2007 40 253.2 254.17 1‐in ARFC, 14‐in doweled JPCP, 4‐in AB 2007 0 0 0
2009/2010 2.7 70 65 0.06 0.0296‐36 1999/2003 101 41 42.5 1‐in ARFC, 11.5‐in non‐dow JPCP, 4‐AB 1999 0 0 0
2009/2010 2.8 54 46 0.04 0.03
98‐159 2001/2003 101 28 34.5 1‐in ARFC, 11.5‐in non‐dow JPCP, 4‐AB 2001 0 0 0
2009/2010 3.6 42 42 0.06 0.04
99‐13 2001/2003 101 24.5 28.25 1‐in ARFC, 11.5‐in non‐dow JPCP, 4‐AB 2001 0 0 0
2009/2010 3 48 45 0.08 0.0397‐68 1999/2003 101 22.69 24.36 1‐in ARFC, 11‐in non‐dow JPCP, 4‐AB 2001 0 0 0
2009/2010 3.7 46 52 0.07 0.0402‐63 2005/2005 202 44.6 47.7 1‐in ARFC, 13‐in non‐dow JPCP, 4‐in AB 2005 0 0 0
2009/2010 1.2 62 57 0.08 0.0203‐06 2005/2005 202 31 33.15 1‐in ARFC, 13‐in non‐dow JPCP, 4‐in ATB 2005 0 0 0
2009/2010 0.4 54 52 0.07 0.01
95‐22 1997/2004 202 13.23 16.44 1‐in ARFC, 12.5‐in non‐dow JPCP, 4‐AB 1997 0 0 0
2009/2010 1.4 38 39 0.05 0.03
92‐39 1994/1994 10 70.66 70.76 0.5‐in ARFC, 13‐in CRCP, 4‐in ATB 1994 0 0 0
2009/2010 17 70 67 0.13 0.117079 1989/2005 101 11.9 14 1‐in ARFC, 9‐in CRCP, 3‐in AC, 6‐in SM 1989 0 0 0
2010 0.7 47 53 0.07 0.02
98
Figure 65. Measured Rutting of ARFC (0.75‐inch) for
Arizona Composite Pavements
Figure 66 shows the correlation between measured and predicted rutting, both of which were
very low, making prediction nearly impossible. The variation of rutting along a project was high
and varied widely from year to year. The MEPDG appears to underpredict rutting by less than
0.10 inch and reasonably predict rutting greater than 0.10 inch.
Figure 66. Measured Versus Predicted Rutting
for ARFC Surfacing Over JPCP
99
Based on these results, the rutting prediction for ARFC over a concrete slab appears to be
reasonable when using the MEPDG models if the proper inputs are used, including an
appropriate binder grade (PG 70‐40) and air voids (12 percent appears to work reasonably well).
Transverse Fatigue Cracking of Existing PCC Slab
The key question is when the ARFC/JPCP is modeled in the MEPDG, and fatigue damage and
cracking are calculated, do they fall in line with the Arizona‐calibrated JPCP curves? The fatigue
damage in the JPCP slab resulting from the millions of heavy truck loadings is computed in the
MEPDG software over the entire analysis period for each section. This fatigue damage is at the
top and bottom of the slab, depending on design and load conditions. If this becomes too high, a
fatigue transverse crack forms between the two transverse joints that define a slab, and a full
width transverse crack develops, leading to spalling, faulting, and, ultimately, slab replacement.
Table 37 summarizes the data used in the transverse fatigue cracking analysis of Arizona
composite pavements. Although these ARFC/JPCP composite pavements were loaded with up to
30 million trucks in the heaviest trafficked lane, fatigue damage to the top and bottom of the
slab was not sufficient to predict any fatigue cracking due to the slab design and the ARFC
surfacing. Field surveys performed on all of these sections also indicated that no midpanel
transverse fatigue cracking existed.
Table 37. Fatigue Damage and Cracking Analysis for
ARFC/JPCP Composite Pavements
Section
ID
Date
PCC/ARFCRoute
Start_
MPEnd_MP Pavement Type
Trucks on
PCC
Millions
Bottom Up
Damage
Top Down
Damage
Measured
Slab
Cracking
Predicted
Slab
Cracking
01‐86 2002/2003 10 254.43 256.1 0.75‐in ARFC, 14.5 in doweled JPCP, 4‐in ATB 0 0.0000 0.0000 0 0
18 0.0000 0.0060 0 0
02‐79 1997/2004 10 157.7 159.2 0.75‐in ARFC, 13‐in doweled JPCP, 3‐in ATB 0 0.0000 0.0000 0 0
30 0.0000 0.0050 0 0
92‐39 1994/1994 10 62 63.48 0.5‐in ARFC, 14‐in doweled JPCP, 4‐in ATB 0 0.0000 0.0000 0 0
20 0.0000 0.0640 0 0
02‐24 2004/2004 10 1.8 11.7 0.75 in ARFC, 13‐in dowled JPCP, AB 0 0.0000 0.0000 0 0
11 0.0000 0.0146 0 0
04‐68 2007/2007 40 253.2 254.17 1‐in ARFC, 14‐in doweled JPCP, 4‐in AB 0 0.0000 0.0000 0 0
5 0.0000 0.0154 0 0
96‐36 1999/2003 101 41 42.5 1‐in ARFC, 11.5‐in non‐dow JPCP, 4‐AB 0 0.0000 0.0000 0 0
6 0.0002 0.0354 0 0.13
98‐159 2001/2003 101 28 34.5 1‐in ARFC, 11.5‐in non‐dow JPCP, 4‐AB 0 0.0000 0.0000 0 0
4.6 0.0000 0.0057 0 0
99‐13 2001/2003 101 24.5 28.25 1‐in ARFC, 11.5‐in non‐dow JPCP, 4‐AB 0 0.0000 0.0000 0 0
3.8 0.0000 0.0045 0 0
97‐68 1999/2003 101 22.69 24.36 1‐in ARFC, 11‐in non‐dow JPCP, 4‐AB 0 0.0000 0.0000 0 0
4.5 0.0000 0.0064 0 0
02‐63 2005/2005 202 44.6 47.7 1‐in ARFC, 13‐in non‐dow JPCP, 4‐in AB 0 0.0000 0.0000 0 0
1.8 0.0000 0.0018 0 0
03‐06 2005/2005 202 31 33.15 1‐in ARFC, 13‐in non‐dow JPCP, 4‐in ATB 0 0.0000 0.0000 0 0
0.6 0.0000 0.0013 0 0
95‐22 1997/2004 202 13.23 16.44 1‐in ARFC, 12.5‐in non‐dow JPCP, 4‐AB 0 0.0000 0.0000 0 0
4.4 0.0000 0.0043 0 0
163 1993/1993 93 SPS‐1 SPS‐1 1 in AC, 15 in RCC 0 0.0000 0.0000 0 0
8.1 0.0000 0.0000 0 0
100
In the analysis the researchers used the Arizona fatigue cracking calibration curve previously
described. This relationship defines concrete slab fatigue damage and transverse midpanel
cracking. Figure 67 shows the results of the analysis (the smooth curve through the data), which
indicate that fatigue cracking can develop as load damage increases over time. The Arizona
calibration curve is defined as:
541
1C
FDICCRK
(Eq. 13)
Where
CRK = low + medium + high cracked slabs as a percentage of all slabs in a mile
DI = Computed Miner’s fatigue damage at the top and bottom of the concrete slab
at the point of maximum tensile stress caused by axle loads and temperature
curl/moisture warping
C4 = 0.19
C5 = ‐2.067
The Arizona bare JPCP sections and 13 Arizona composite sections are plotted in Figure 67. The
data fall right in line with the JPCP sections from Arizona. None of the Arizona ARFC/JPCP
composite pavements showed transverse fatigue cracking and none was predicted by the
MEPDG because of the relatively thick concrete slabs and the ARFC surfacing that reduces
thermal and moisture gradients within the slab, thus limiting slab curling and fatigue damage.
These results in Figure 67 indicate that the ARFC/JPCP composite pavements fatigue damage
and cracking are analyzed reasonably well in the MEPDG and that the Arizona calibrated PCC
fatigue models are applicable. The researchers concluded that the Arizona JPCP transverse
fatigue cracking damage model holds reasonably well for Arizona based on these composite
project results.
101
Figure 67. Percentage of Slabs Cracked Versus JPCP Slab Fatigue Damage in
JPCP LTPP and PMS Sections and in Composite Pavement
IRI (Smoothness)
The IRI model for flexible pavement is used for composite pavement. The predicted IRI depends
mainly on the initial IRI after construction and developing distresses such as rutting and
cracking. The researchers determined the initial IRI values from the ADOT PMS database for the
closest year after construction; if this information was unavailable, the researchers backcasted
from future available measurements. The mean initial IRI after construction for ARFC/JPCP was
53 inches/mi (with a range of 36 to 65 inches/mi), which is very smooth.
Other factors include slab cracking, transverse joint reflection as well as surface rutting. SF is
also included with freezing depth, time since construction, and type of subgrade soil. Since
rutting and transverse cracking were very low for ARFC/JPCP, the future IRI depends on the
initial IRI. If transverse joint reflection cracking developed within a few years, that too would
affect IRI in the MEPDG.
Figure 68 plots predicted versus measured IRI. The flexible pavement IRI model appears to
reasonably predict the IRI of composite ARFC/JPCP projects over time and traffic with no further
modification to the calibration coefficients.
102
Figure 68. MEPDG Predicted Versus Measured IRI for
Composite Pavement Projects (2 to 16 Years)
Joint Reflection Cracking
Field surveys of all ARFC/JPCP sections showed that all transverse joints had reflected through.
These reflected joints ranged in age from 7 to 16 years and were mostly under very heavy truck
traffic. Nearly all joints showed a low severity deterioration; only one project showed medium
severity. The MEPDG empirical model predicts that all of these joints will reflect through within
a few years, which occurred. However, a large majority of these reflection cracks did not
deteriorate and affect rideability. They are still expected to deteriorate, especially nondoweled
transverse joints, which will lead to ARFC replacement.
These results indicate that designers can use the MEPDG to design ARFC/JPCP composite
pavements with confidence. The HMA overlay of JPCP mode must be selected to design a
composite pavement.
CRCP and ARFC/CRCP Composite
Arizona has constructed two CRCP sections over the years that are available for use in distress
and IRI model validation: Loop 101, Milepost (MP) 8 to MP 14.9 (northbound lane), and I‐10, MP
103
70.66 to MP 70.76 (westbound lane), described below. The researchers collected data for these
sections through field surveys, ADOT files, and the LTPP database.
Loop 101, MP 8 to 14.9 (between Northern and Bell roads), LTPP section 11.9‐12.0
(northbound lane). Built in 1989, this section was bare concrete until 2005 when a 0.75‐inch
ARFC was placed on Phoenix‐area freeways to cut noise level. Other details of this section are:
9‐inch CRCP, 3‐inch AC base, A‐6 subgrade.
Longitudinal steel: #5 bars, spaced at 6 inches and placed at mid‐depth with
0.568 percent reinforcement.
Performance:
o CRCP exhibited no distress in 2005 when it was resurfaced with 0.75‐inch ARFC.
o Crack spacing: 40 inches. All cracks were very tight.
o The 2012 survey showed that no transverse cracks had reflected through the
entire length of thin ARFC after seven years of heavy traffic. The ARFC exhibited
no distress and was in excellent condition.
I‐10, MP 70.66 to 70.76, (westbound lane). This very short section was constructed in 1994 as
an experimental section and was immediately surfaced with a very thin 0.5‐inch ARFC. Other
details of this section follow:
0.5‐inch ARFC at construction.
13‐inch CRCP, 6‐inch aggregate base, A‐2‐4 subgrade.
AC shoulder.
Longitudinal steel: #6 bars, spaced at 6 inches and placed at mid‐depth with
0.57 percent reinforcement.
Performance:
o Mean reflected crack spacing was about 48 to 60 inches in early 2011. Some of
these cracks were about 0.10 inch wide and others were tight.
o The CRCP exhibited two medium/high punchouts in January 2011 per 528 ft,
which translates to 20 per mile.
o The ARFC exhibited raveling distress and was in fair to poor condition after 16
years of service under heavy truck traffic.
104
Inputs for the DARWin ME were obtained for these two sections as well as their performance
through early 2011. The predicted performance matched the field performance well, as
summarized in Table 38. These results suggest that the MEPDG is predicting well the
performance of these two Arizona CRCP projects over long service lives and heavy traffic. The
Loop 101 CRCP has shown excellent performance over 21 years (tight transverse cracks, no
punchouts, smooth), and the MEPDG predicts excellent performance in every way. The I‐10
ARFC (0.5 inch)/CRCP has been under much heavier traffic (20 million trucks in the outer lane)
and has recently developed wide cracks and structural punchouts. The MEPDG predicts this
deterioration as shown by the predicted wide cracks, loss of crack load transfer efficiency, and
punchouts. If the I‐10 ARFC/CRCP had been designed with a higher steel percentage and an AC
base, the pavement would have performed very well over the design period under more than
25 million truck loadings. The ARFC of 0.5 inch is too thin to have any beneficial effect.
Table 38. Measured and Predicted Performance of Loop 101 and I‐10 CRCP Sections
Performance
Indicator
Loop 101, MP 8 to MP 14.9 NB I‐10, MP 70.66 to MP 70.76 WB
Measured Predicted Measured Predicted
Crack spacing 40 inches 44 inches
48‐ to 60‐inch
mean
54 inches
Crack width Could not
measure, but
all very tight
(<0.020 inches)
<0.020 inches
Could not
measure,
but some
>0.040 inches
0.040 inches
(very wide; would
cause loss of load
transfer efficiency)
Crack load transfer
efficiency Very high,
90% to 100% >92% to 98% 40% to 100%
Could not measure,
but some wide and
poor (<60%)
Punchouts (M, H) 0 per mile 0 per mile 20 per mile 5 per mile
IRI 47 inch/mi* 53 inch/mi* 90 inch/mi 79 inch/mi
Number of trucks 7 million outer lane 20 million outer lane
Time period 1989 to 2010 (21 years) 1994 to 2010 (16 years)
*Measured and predicted on ARFC surface five years after it was placed.
Does the performance of the two Arizona CRCP sections match the Arizona calibration for CRCP
conducted under NCHRP 20‐07 (calibration done for corrected CTE values)? The fatigue damage
in the CRCP slab resulting from the millions of heavy truck loadings is computed in the MEPDG
software over the entire analysis period. This fatigue damage is at the top of the slab, 48 inches
from the slab edge. If this becomes too high, a fatigue longitudinal crack forms between two
transverse cracks, initiating the development of an edge punchout. The NCHRP 20‐07 national
105
calibration of the CRCP punchout model developed a relationship between concrete fatigue
damage at the top of the slab and punchouts. This plot, shown in Figure 69 (the smooth curve
through the data), shows how the development of structural punchouts can form as load
damage increases over time.
Figure 69. CRCP Observed Punchouts Versus Fatigue Damage for
All CRCP Used in the 2007 National Calibration
(Two Arizona CRCP sections circled in red)
The national calibration curve is defined as:
541
3CDAMAGEC
CPO
(Eq. 14)
I‐10
Loop 101
106
Where
PO = number of medium and high punchouts per mile
DAMAGE = computed Miner’s fatigue damage at top of concrete slab at
point of maximum tensile stress caused by axle loads and
temperature curl/moisture warping
C3 = 85
C4 = 1.4149
C5 = ‐0.8061
The Loop 101 and I‐10 sections are also plotted in Figure 69. These points are within the same
general grouping of U.S. CRCP sections. Loop 101 CRCP is nowhere close to developing
punchouts but I‐10 CRCP has developed some. While this is very limited CRCP performance data,
the researchers concluded that the national CRCP punchout damage model holds reasonably
well for Arizona based on the results of these two projects. CRCP may be designed in Arizona
using the national calibration factors along with the standard deviation for reliability design.
107
CHAPTER 4. SENSITIVITY ANALYSIS
To help validate the local Arizona model predictions, the researchers performed a
comprehensive, 30‐year sensitivity analysis of the new HMA and new JPCP models. The
researchers could have chosen a much longer time period, such as 50 years, however the
relative results were not expected to change. They developed new HMA pavement and JPCP
baseline designs with selected inputs to represent typical Arizona site conditions, design
features, construction practices, and pavement materials. Details of the baseline designs for
new HMA and new JPCP are presented in Tables 39 and 40, respectively. Key inputs were varied
one at a time except where two inputs have known correlations, such as PCC thickness and
dowel size. Dowel diameter was increased with PCC thickness using the thickness divided by 8
ratio. Thus, a 12‐inch slab would have a 1.50‐inch dowel diameter. Next, the researchers plotted
the predicted outcomes for each input to illustrate their impact on cracking, faulting, and IRI
smoothness, and assessed the reasonableness of the impact on these key performance outputs.
Table 39. Key Inputs in New or Reconstructed HMA Pavement Sensitivity Analysis
Input Parameter Values
Lower End Mean (Baseline) Upper End
HMA thickness (inches) 3 6 9
Air voids (%) 6 7 8
Binder content (%) 8 11 14
Binder type (Superpave) 64‐22 70‐22 76‐16
Subgrade (AASHTO soil class) A‐2‐4 A‐6 —
Initial average annual daily
truck traffic (AADTT) 2000 6000 10,000
Climate (weather stations) Kingman, Page St. Johns Douglas, Bisbee
Reliability (%) 50 75 95
108
Table 40. Key Inputs in New or Reconstructed JPCP Sensitivity Analysis*
Input Parameter Values
Lower End Mean (Baseline) Upper End
PCC thickness (inches) 8 10 12
CTE (inch/inch/° F) 4.5 5.0 5.5
AADTT 5000 10,000 15,000
Dowel diameter (inches)
(used PCC thickness/8 rule) 1 1.25 1.5
Subgrade (AASHTO soil class) A‐6 A‐2‐4 —
Shoulder type HMA Tied PCC —
Climate (weather stations) Tucson Phoenix Kingman
Reliability (%) 50 75 95
*Similar results would be obtained for ARFC/JPCP.
Results of the JPCP sensitivity analysis are presented in the following sections. The sensitivity
analysis indicates reasonable results that fairly well match the results obtained in field
experiments (e.g., thicker slabs show less cracking; larger dowels show less faulting).
NEW HMA SENSITIVITY ANALYSIS RESULTS
Effect of Traffic
Figures 70 through 72 show the effect of average annual daily truck traffic (AADTT) on predicted
distress and IRI. The initial year two‐directional AADTT varied from 2000 to 10,000, which
represents a large range of very heavy truck traffic. The 2000 AADTT results in about 10 million
total trucks in the design lane over 20 years. The 10,000 AADTT results in about 48 million trucks
in the design lane over 20 years.
The trends of predicted distress and IRI were as expected, with increasing traffic applications
resulting in higher levels of predicted distress and IRI. Although the trends were all reasonable,
the effect of AADTT on predicted IRI was not significant. These results are in accordance with
generally accepted HMA performance in the field.
109
0
20
40
60
80
100
0 5 10 15 20
Alligator Cracking
, Percent Lan
e Area
Pavement Age, Years
AADTT 2000 AADTT 6000 (Baseline) AADTT 10000
Figure 70. Effect of AADTT on HMA Alligator Fatigue Cracking
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20
Total R
utting
, in
Pavement Age, Years
AADTT 2000 AADTT 6000 (Baseline) AADTT 10000
Figure 71. Effect of AADTT on HMA Total Rutting
110
40
80
120
160
200
0 5 10 15 20
IRI, in/m
ile
Pavement Age, Years
AADTT 2000 AADTT 6000 (Baseline) AADTT 10000
Figure 72. Effect of Initial AADTT on HMA IRI
Effect of HMA Thickness
Figures 73 through 75 show the effect of HMA thickness on predicted distress and IRI. Thicker
HMA provides increased structure for applied traffic loads, reduces stresses and the amount of
fatigue damage at the bottom of the HMA layer, and also reduces HMA layer deflections.
Therefore, increased HMA thickness should result in less alligator cracking, rutting, and IRI.
The three plots show that HMA thickness had the most significant effect on alligator cracking
and rutting. Increased HMA thickness resulted in lower alligator cracking and rutting, and had a
smaller but significant effect on IRI reduction. These trends are typical and as expected when
compared to field pavement performance data from in‐service flexible pavements.
111
0
20
40
60
80
100
0 5 10 15 20
Alligator Cracking
, Percent Lan
e Area
Pavement Age, Years
3 in AC 6 in AC (Baseline) 9 in AC
Figure 73. Effect of HMA Thickness on HMA Alligator Fatigue Cracking
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20
Total R
utting
, in
Pavement Age, Years
3 in AC 6 in AC (Baseline) 9 in AC
Figure 74. Effect of HMA Thickness on Total Rutting
112
40
80
120
160
200
0 5 10 15 20
IRI, in/m
ile
Pavement Age, Years
3 in AC 6 in AC (Baseline) 9 in AC
Figure 75. Effect of HMA Thickness on IRI
Effect of HMA Binder Content
Figures 76 through 78 show the effect of HMA binder content on predicted distress and IRI.
Higher HMA binder content reduces tensile strain and the amount of fatigue damage at the
bottom of the HMA layer thereby reducing alligator cracking. Increased binder content
increased rutting as shown. However, binder content has minimal effect on IRI.
The three plots show that HMA binder content had the most significant effect on alligator
cracking. Increased binder content resulted in lower alligator cracking and higher rutting. HMA
binder content had a small effect on IRI. These trends are typical and as expected when
compared to field pavement performance data from in‐service flexible pavements.
113
0
20
40
60
80
100
0 5 10 15 20
Alligator C
racking, Percent Lan
e Area
Pavement Age, Years
8% Binder 11% Binder (Baseline) 14% Binder
Figure 76. Effect of Binder Content on HMA Alligator Cracking
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20
Total R
uttin
g, in
Pavement Age, Years
8% Binder 11% Binder (Baseline) 14% Binder
Figure 77. Effect of Binder Content on HMA Total Rutting
114
40
80
120
160
200
0 5 10 15 20
IRI, in/m
ile
Pavement Age, Years
8% Binder 11% Binder (Baseline) 14% Binder
Figure 78. Effect of Binder Content on HMA IRI
Effect of HMA Air Voids
Figures 79 through 81 show the effect of HMA air voids on predicted distress and IRI. Lower
HMA air voids increases the dynamic modulus and reduces strain. Therefore, the amount of
fatigue cracking at the bottom of the HMA layer is reduced. Decreased air voids have a lower
but significant effect on reducing rutting. However, air voids has minimal effect on IRI.
The three plots show that HMA air voids had the most significant effect on alligator cracking.
Decreased air voids resulted in lower alligator cracking and rutting. Air voids has minimal effect
on IRI. These trends are typical and as expected when compared to field pavement performance
data from in‐service flexible pavements.
115
0
20
40
60
80
100
0 5 10 15 20
Alligator Cracking
, Percent Lan
e Area
Pavement Age, Years
6% Air Voids 7% Air Voids (Baseline) 8% Air Voids
Figure 79. Effect of Air Voids on HMA Alligator Cracking
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20
Total R
utting
, in
Pavement Age, Years
6% Air Voids 7% Air Voids (Baseline) 8% Air Voids
Figure 80. Effect of Air Voids on HMA Total Rutting
116
40
80
120
160
200
0 5 10 15 20
IRI, in/m
ile
Pavement Age, Years
6% Air Voids 7% Air Voids (Baseline) 8% Air Voids
Figure 81. Effect of Air Voids on HMA IRI
Effect of HMA Binder Grade
Figures 82 through 84 show the effect of HMA PG binder on predicted distress and IRI. Higher
PG binder reduces both alligator cracking and rutting.
The three plots show that HMA binder grade had the most significant effect on alligator cracking
and rutting. For both alligator cracking and rutting performance measures, increased binder
grade resulted in lower alligator cracking and rutting. These trends are typical and as expected
when compared to field pavement performance data from in‐service flexible pavements.
117
0
20
40
60
80
100
0 5 10 15 20
Alligator Cracking
, Percent Lan
e Area
Pavement Age, Years
PG64‐22 PG70‐22 (Baseline) PG76‐16
Figure 82. Effect of PG Binder on HMA Alligator Cracking
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20
Total R
utting
, in
Pavement Age, Years
PG64‐22 PG70‐22 (Baseline) PG76‐16
Figure 83. Effect of PG Binder on HMA Total Rutting
118
40
80
120
160
200
0 5 10 15 20
IRI, in/m
ile
Pavement Age, Years
PG64‐22 PG70‐22 (Baseline) PG76‐16
Figure 84. Effect of PG Binder on HMA IRI
Effect of Subgrade Type
Figures 85 through 87 show the effect of subgrade type on predicted distress and IRI. Subgrade
support is considered in DARWin‐ME through the soil’s resilient modulus (at optimum moisture
content and density). Typical default mean resilient modulus values were used for these two
soils: A‐2‐4 (coarse‐grain subgrade materials) was the highest and A‐6 (fine‐grain materials) the
lowest. The temperature and moisture models used also act upon the soils on a monthly basis,
producing varying values. In general, as the resilient modulus increases, load‐associated stresses
decrease. The three plots show that subgrade type had a small effect on all three performance
measures, with A‐2‐4 performing slightly better than the weaker A‐6 materials.
119
0
20
40
60
80
100
0 5 10 15 20
Alligator Cracking
, Percent Lan
e Area
Pavement Age, Years
A‐6 (Baseline) A‐2‐4
Figure 85. Effect of Subgrade Type on HMA Alligator Cracking
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20
Total R
utting
, in
Pavement Age, Years
A‐6 (Baseline) A‐2‐4
Figure 86. Effect of Subgrade Type on HMA Total Rutting
120
40
80
120
160
200
0 5 10 15 20
IRI, in/m
ile
Pavement Age, Years
A‐6 (Baseline) A‐2‐4
Figure 87. Effect of Subgrade Type on HMA IRI
Effect of Climate
Figures 88 through 90 show the effect of climate on predicted distress and IRI. Results from the
Phoenix area were added to these plots to increase the variation in weather. Table 41
summarizes the relationship between elevation, cracking, and rutting. Transverse cracking
results from Chapter 3 are also included.
121
0
20
40
60
80
100
0 5 10 15 20
Alligator Cracking
, Percent Lan
e Area
Pavement Age, Years
St. Johns (Baseline) Page Kingman Douglas Bisbee Phoenix
Figure 88. Effect of Climate on HMA Alligator Cracking
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20
Total R
utting
, in
Pavement Age, Years
St. Johns (Baseline) Page Kingman Douglas Bisbee Phoenix
Figure 89. Effect of Climate on HMA Total Rutting
122
40
80
120
160
200
0 5 10 15 20
IRI, in/m
ile
Pavement Age, Years
St. Johns (Baseline) Page Kingman Douglas Bisbee Phoenix
Figure 90. Effect of Climate on HMA IRI
Table 41. Effect of Climate on Predicted Distress and IRI
Site Location Elevation
(ft) Rutting (inches)
AlligatorCracking
(%)
Transverse Cracking
(ft)
St. Johns 5722 0.52 11 2112
Bisbee 4105 0.63 20 1966
Kingman 3420 0.60 15 1831
Phoenix 1105 0.85 40 0
Climate varies with elevation, which results in varying amounts of fatigue cracking, rutting, and
low‐temperature cracking. Generally, as elevation increases, rutting and fatigue cracking
decrease and low‐temperature transverse cracking increases. These results can be explained
partially through temperature considerations and their impact on the HMA dynamic modulus
E*. Higher elevations have colder temperatures and higher E*, which reduces fatigue damage
123
and permanent deformation. However, low‐temperature cracking becomes more significant in
higher elevations.
The trends reported in Figures 88 through 90 are typical and as expected when compared to
field pavement performance data from in‐service HMA located in different climates.
Effect of Reliability
Figures 91 through 93 show the effect of reliability on predicted distress and IRI. A 50 percent
design reliability indicates that the alligator cracking, rutting, and IRI curve ARE being predicted
on average. Thus, half of the projects designed should show higher alligator cracking, rutting,
and IRI, and half should show lower values.
As the design reliability increases to the 95 percent level, the 95 percent curve increases,
indicating that 95 of 100 HMA sections would show this amount or less of alligator cracking,
rutting, and IRI. Note that the standard deviation associated with calculating the reliability
curves are based on the SEE of the calibration data associated with models. The higher the
error, the greater the spread of these curves from 50 to 95 percent for any example and thus
the higher the required HMA thickness.
The results show that the design reliability curve for 95 percent is higher than any lower level of
reliability. At higher and higher levels of reliability, a thicker pavement or higher support base
course will be needed to meet the design reliability requirements.
Figure 91. Effect of Reliability on HMA Alligator Cracking
124
Figure 92. Effect of Reliability on HMA Total Rutting
Figure 93. Effect of Reliability on HMA IRI
125
Overall these analyses suggest that the MEPDG model prediction is consistent in prediction of
distress and IRI at least in the same direction as theory and previous experience and field
performance have shown. Changes in the design inputs affect HMA alligator cracking, rutting,
and IRI smoothness by varying amounts. Knowledge of this cause and effect is useful to design
engineers in developing cost‐effective pavements.
NEW JPCP SENSITIVITY ANALYSIS RESULTS
Effect of Traffic
Figures 94 through 96 show the effect of AADTT on predicted distress and IRI. The initial year
AADTT varied from 5000 to 15,000, which represented a large range of very heavy truck traffic.
The 5000 daily trucks results in about 35 million total trucks in the design lane over 30 years;
15,000 daily trucks results in about 104 million trucks in the design lane over 30 years.
The results of this analysis show that, in all cases, the increased number of heavy trucks results
in greater slab fatigue cracking, joint faulting, and IRI. These results are in accordance with
generally accepted JPCP performance in the field.
0
20
40
60
80
100
0 5 10 15 20 25 30
Percen
t Slab
s Cracked
Pavement Age, Years
AADTT 5000 AADTT 10000 (Baseline) AADTT 15000
Figure 94. Effect of AADTT on JPCP Transverse Cracking
126
0.0
0.3
0.6
0.9
1.2
1.5
0 5 10 15 20 25 30
Mean Joint Fau
lting, in
Pavement Age, Years
AADTT 5000 AADTT 10000 (Baseline) AADTT 15000
Figure 95. Effect of AADTT on JPCP Mean Joint Faulting
60
160
260
360
460
560
660
760
0 5 10 15 20 25 30
IRI, in/m
ile
Pavement Age, Years
AADTT 5000 AADTT 10000 (Baseline) AADTT 15000
Figure 96. Effect of Initial AADTT on JPCP IRI
127
Effect of PCC Thickness
Figures 97 through 99 show the effect of PCC thickness on predicted distress and IRI. Thicker
PCC slabs provide increased structure for applied traffic loads and thus reduce midpanel stresses
and the amount of fatigue damage at the top and bottom of the slab, and deflections at the slab
corners and transverse joints causing less erosion. Therefore, increased PCC slab thickness
should result in less cracking, faulting, and IRI.
The three plots show that slab thickness had the most significant effect on cracking and IRI. For
these performance measures, increased thickness resulted in lower cracking and IRI. For joint
faulting, increasing PCC thickness had a smaller but significant effect on reduced faulting. These
trends are typical and as expected when compared to field pavement performance data from in‐
service JPCP.
0
20
40
60
80
100
0 5 10 15 20 25 30
Percen
t Slab
s Cracked
Pavement Age, Years
8 in PCC 10 in PCC (Baseline) 12 in PCC
Figure 97. Effect of PCC Thickness on JPCP Transverse Cracking
128
0.0
0.3
0.6
0.9
1.2
1.5
0 5 10 15 20 25 30
Mean Joint Fau
lting, in
Pavement Age, Years
8 in PCC 10 in PCC (Baseline) 12 in PCC
Figure 98. Effect of PCC Thickness on JPCP Mean Joint Faulting
60
160
260
360
460
560
660
760
0 5 10 15 20 25 30
IRI, in/m
ile
Pavement Age, Years
8 in PCC 10 in PCC (Baseline) 12 in PCC
Figure 99. Effect of PCC Thickness on JPCP IRI
129
Effect of CTE
Figures 100 through 102 show the effect of CTE on predicted distress and IRI. CTE impacts the
change in length of concrete when a temperature difference is applied. It affects slab curling and
joint opening/closing, thus having an impact on joint faulting and transverse cracking. DARWin‐
ME is the first pavement design procedure to consider the impact of CTE.
The three plots show that PCC CTE has a very significant effect on all three performance
measures. Higher CTE results in higher levels of pavement deterioration in all cases. This result is
expected, although little independent field data are available to support such an effect. As CTE
influences the magnitude of stresses and deflections/curling of the PCC slab, an increased CTE
would be expected to increase stresses and deflections, which normally would result in more
cracking and faulting, and thus higher IRI. The increase in pavement life (when IRI reaches
150 inches/mi) is from 4 years to more than 30 years with a reduction in CTE of 5.5 to 4.5×10‐6,
which could be caused by a change in coarse aggregate.
0
20
40
60
80
100
0 5 10 15 20 25 30
Percen
t Slab
s Cracked
Pavement Age, Years
4.5 CTE 5.0 CTE (Baseline) 5.5 CTE
Figure 100. Effect of CTE on JPCP Transverse Cracking
130
0.0
0.3
0.6
0.9
1.2
1.5
0 5 10 15 20 25 30
Mean Joint Fau
lting, in
Pavement Age, Years
4.5 CTE 5.0 CTE (Baseline) 5.5 CTE
Figure 101. Effect of CTE on JPCP Mean Joint Faulting
60
160
260
360
460
560
660
760
0 5 10 15 20 25 30
IRI, in/m
ile
Pavement Age, Years
4.5 CTE 5.0 CTE (Baseline) 5.5 CTE
Figure 102. Effect of CTE on JPCP IRI
131
Effect of Shoulder Type
Figures 103 through 105 show the effect of shoulder type on predicted distress and IRI.
DARWin‐ME models show that the support provided at the slab and shoulder interface has a
significant impact on joint deflections and critical stresses midpanel at the bottom of the PCC
slab. For JPCP, support provided at the slab and shoulder interface edge support is highly
influenced by the type of shoulder provided. Typically, the best edge support is provided by tied
PCC shoulders with high load transfer efficiency. This is followed by PCC shoulders (not tied) and
AC shoulders that provide similar levels of support. Note that the longevity of the support
provided also has a considerable impact on pavement performance.
Figures 103 through 105 show the effect of shoulder type on predicted cracking, faulting, and
IRI, respectively. For all three performance measures, provision of tied PCC shoulders had a
positive impact on performance (i.e., reduced cracking, faulting, and IRI). The impact of shoulder
type on distress and IRI was most significant for cracking. For both faulting and IRI, the impact
was less significant. These trends are as expected when compared to field pavement
performance data from in‐service JPCP with various shoulder types.
0
20
40
60
80
100
0 5 10 15 20 25 30
Percen
t Slab
s Cracked
Pavement Age, Years
AC Shoulder Tied PCC Shoulder (Baseline)
Figure 103. Effect of Shoulder Type on JPCP Transverse Cracking
132
0.0
0.3
0.6
0.9
1.2
1.5
0 5 10 15 20 25 30
Mean Joint Fau
lting, in
Pavement Age, Years
AC Shoulder Tied PCC Shoulder (Baseline)
Figure 104. Effect of Shoulder Type on JPCP Mean Joint Faulting
60
160
260
360
460
560
660
760
0 5 10 15 20 25 30
IRI, in/m
ile
Pavement Age, Years
AC Shoulder Tied PCC Shoulder (Baseline)
Figure 105. Effect of Shoulder Type on JPCP IRI
133
Effect of Subgrade Type
Figures 106 through 108 show the effect of subgrade type on predicted distress and IRI.
Subgrade support is considered in DARWin‐ME through the resilient modulus (at optimum
moisture content and density) of the soil. Typical default mean resilient modulus values were
used for these two soils, with the coarse‐grained subgrade materials (A‐2‐4) being the highest
and the fine‐grained (A‐6) soil the lowest. The temperature and moisture models of these soils
also act upon the soils on a monthly basis, producing different values. In general, as the resilient
modulus increases, load‐associated stresses decrease, but the thermal curling stress increases.
The relative values of these two causes of stress dominate for different designs and climates and
result in differing levels of stress (and, thus, different levels of cracking and joint faulting).
The three plots show that subgrade type had a small effect on cracking but more significant
effect on joint faulting and IRI. The A‐2‐4 soil showed lower joint faulting and IRI performance
than the A‐6 soil. These trends are typical and as expected when compared to field pavement
performance data from in‐service JPCP.
0
20
40
60
80
100
0 5 10 15 20 25 30
Percen
t Slab
s Cracked
Pavement Age, Years
A‐6 A‐2‐4 (Baseline)
Figure 106. Effect of Subgrade Type on JPCP Transverse Cracking
134
0.0
0.3
0.6
0.9
1.2
1.5
0 5 10 15 20 25 30
Mean Joint Fau
lting, in
Pavement Age, Years
A‐6 A‐2‐4 (Baseline)
Figure 107. Effect of Subgrade Type on JPCP Mean Joint Faulting
60
160
260
360
460
560
660
760
0 5 10 15 20 25 30
IRI, in/m
ile
Pavement Age, Years
A‐6 A‐2‐4 (Baseline)
Figure 108. Effect of Subgrade Type on JPCP IRI
135
Effect of Climate
Figures 109 through 111 show the effect of climate on predicted distress and IRI. Climate ranges
from Phoenix at an elevation of 1105 ft to Tucson at 2400 ft to Kingman at 3420 ft. Transverse
fatigue cracking is significantly affected with the highest amount occurring in the Kingman
climate due to an increased amount of large temperature gradients through the slab as
compared to the warmer desert climates.
The figures show that climate has the most significant effect on cracking where the high
elevation desert climate (represented by Kingman) was most detrimental due to large
temperature gradients and low humidity. Climate had a minimal effect on doweled joint faulting
and IRI. These trends are typical and as expected when compared to field pavement
performance data from in‐service JPCP located in different climates.
0
20
40
60
80
100
0 5 10 15 20 25 30
Percen
t Slab
s Cracked
Pavement Age, Years
Kingman Tucson Phoenix (Baseline)
Figure 109. Effect of Climate on JPCP Transverse Cracking
136
0.0
0.3
0.6
0.9
1.2
1.5
0 5 10 15 20 25 30
Mean Joint Fau
lting, in
Pavement Age, Years
Kingman Tucson Phoenix (Baseline)
Figure 110. Effect of Climate on JPCP Mean Joint Faulting
60
160
260
360
460
560
660
760
0 5 10 15 20 25 30
IRI, in/m
ile
Pavement Age, Years
Kingman Tucson Phoenix (Baseline)
Figure 111. Effect of Climate on JPCP IRI
137
Effect of Reliability
Figures 112 through 114 show the effect of reliability on predicted distress and IRI. A level of
design reliability of 50 percent indicates that the mean slab cracking, joint faulting, and IRI curve
are being predicted. Thus, half of the projects designed should show higher cracking, faulting,
and IRI, and half should show lower values.
As the design reliability increases to the 95 percent level, the 95 percent curve increases,
indicating that 95 of 100 JPCP sections would show this amount or less of cracking, faulting, and
IRI. The standard deviation associated with calculating the reliability curves is based on the SEE
of the calibration data associated with models. The higher the error, the greater the spread of
these curves from 50 to 95 percent for any example.
The results show that the design reliability curve for 95 percent is higher than any lower level of
reliability. The consequence for design at higher and higher levels of reliability is that a thicker
pavement, larger dowels, or higher support base course will be needed to meet the design
reliability requirements.
An overall review of the sensitivity analysis results indicates that DARWin‐ME model prediction
is consistent with expected trends from empirical observations and pavement engineering
theory. The sensitivity analysis provides information about the inputs that have the most impact
on pavement deterioration, which this information is useful to design engineers for developing
cost‐effective designs.
Figure 112. Effect of Reliability on JPCP Transverse Cracking
138
Figure 113. Effect of Reliability on JPCP Mean Joint Faulting
Figure 114. Effect of Reliability on JPCP IRI
139
CHAPTER 5. VALIDATION OF LOCALLY CALIBRATED DARWin‐ME
NEW HMA AND JPCP DESIGN PROCEDURES FOR ARIZONA
The Arizona locally calibrated DARWin‐ME new HMA and JPCP design procedures were further
validated through direct comparison of designs obtained using the Materials Preliminary
Engineering and Design Manual (ADOT 1989) and the locally calibrated ADOT DARWin‐ME. Since
DARWin‐ME is a far more efficient design tool regarding design optimization, some differences
are expected in designs developed using the two different procedures. However, the differences
are not expected to be extremely significant.
Pavement designs were developed for two typical ADOT HMA and JPCP projects (obtained from
the calibration database). Before comparing the designs, they were evaluated for
reasonableness (i.e., if the designs had performed as expected) as follows:
Performed as expected (i.e., little or no indications of distress after 20 to 30 years of
service). For these pavements, new designs developed by the ADOT Pavement Design
Guide and ADOT DARWin‐ME are expected to be similar to the existing in‐service
pavements.
Exhibited early failures (i.e., high levels of distress within 10 to 15 years of a 20‐ to 30‐
year design). For these pavements, new designs developed by the ADOT Pavement
Design Guide and ADOT DARWin‐ME are expected to be an improvement over the
existing pavement design (e.g., thicker HMA and PCC layers).
Results are presented in the remainder of this chapter.
NEW HMA PAVEMENT DESIGN COMPARISON
Project Description
Inputs for the new HMA pavement design comparisons were adapted from a project on I‐40
near Holbrook, Arizona. This project has HMA over a 10‐inch aggregate base placed over an A‐6
subgrade.
For this reconstructed project, the researchers established a trial design based on the site,
structure, materials, and other properties of the existing pavement (Figure 115). A design period
of 20 years was selected based on current ADOT new HMA pavement design practices. Table 42
presents key inputs for both the Materials Preliminary Engineering and Design Manual (ADOT
1989) and the locally calibrated ADOT DARWin‐ME.
140
Figure 115. New HMA Project Trial Design Structure
(TBD? = to be determined during design)
141
Table 42. Key Inputs for New HMA Pavement Project Design Comparison
Input Category
Input Variables
Inputs for MaterialsPreliminary
Engineering and Design Manual
Inputs for Locally Calibrated ADOT DARWin‐ME
Structure and
materials type
and
properties
HMA ADOT AC ½‐inch mix
(Layer coefficient: 0.44)
Binder grade: PG 70‐22
Binder content: 11.3%
Air voids: 6%
Aggregate base
Layer coefficient: 0.14
Drainage coefficients:
0.9
AASHTO Soil Class: A‐1‐a
Aggregate base resilient modulus:
34,000 psi
Subgrade
(AASHTO Soil A‐6)
Subgrade resilient
modulus @ in situ
moisture content
(8000 psi)
Subgrade resilient modulus @
optimum moisture content
(14900 psi)*
Design Design period 20 years 20 years
Design reliability 97% 97%
Traffic
Initial AADTT 7200 7200
Directional distribution 0.5 0.5
Lane distribution 0.8 0.8
Truck growth 3% compound 3% compound
Cumulative ESAL 37.76 million —
Cumulative trucks — 28.27 million
Performance
criteria
Design reliability 97% 97%
Initial serviceability 4.2 —
Terminal serviceability 3.0 —
Overall standard
deviation 0.35 —
Initial IRI (inch/mi) — 40
Terminal IRI (inch/mi) — 150
Alligator cracking (%) — 10
Total rutting (inch) — 0.55
*The laboratory resilient modulus tests are performed on representative samples in stress and moisture conditions
simulating those of the primary moisture seasons (AASHTO 1993). On the other hand, MEPDG requires resilient
modulus at optimum moisture and maximum dry density conditions. Therefore, a correction was applied to the
laboratory optimum resilient modulus to estimate resilient modulus at field in situ conditions.
Design Analysis
The researchers used DARWin 3.1 software (software version of the AASHTO Guide for Design of
Pavement Structures [1993]) and AASHTO DARWin‐ME in the design analysis. Both software
inputs and calibration coefficients were modified as needed to be compatible with the Materials
142
Preliminary Engineering and Design Manual (ADOT 1989) and the locally calibrated pavement
models. The trial designs were run for 20 years. The results are presented in Table 43. Figures
116 through 118 show predictions of alligator cracking, total rutting, and IRI from the local
DARWin M‐E models.
The locally calibrated DARWin‐ME procedure produced a thinner pavement section (HMA
thickness of 12.25 inch) when compared to the current ADOT pavement design procedure (HMA
thickness = 11.0 inch). Based on the design produced, the researchers concluded that the locally
calibrated DARWin‐ME models and design procedure produce reasonable new HMA pavement
designs for this high level of truck traffic.
Table 43. Trial Design Results for New HMA Pavement Projects
Key Variables Materials Preliminary
Engineering and Design Manual Locally Calibrated ADOT
DARWin‐ME
Required structural number 6.66 —
Design equivalent single axle
loads (ESALs) 37.76 million 37.76 million
Cumulative trucks — 28.27 million
Reliability 97% 97%
Subgrade R‐value (MR) 8000 psi 14,900 psi
Aggregate base thickness 10 inches 10 inches
Design HMA thickness 12.25 inches 11 inches
Figure 116. Predicted Alligator Cracking for the New HMA Trial Design Using DARWin‐ME
143
Figure 117. Predicted Rutting for the New HMA Trial Design Using DARWin‐ME
Figure 118. Predicted IRI for the New HMA Trial Design Using DARWin‐ME
144
NEW JPCP DESIGN COMPARISON
Project Description
Inputs for the new JPCP design comparisons were adapted for a project on I‐10 west of Phoenix.
The section consists of JPCP over a 6‐inch aggregate base placed over an A‐2‐4 subgrade.
For the redesign of this project, the researchers also established a trial design based on the site,
structure, materials, and other properties of the existing pavement (Figure 119). A 30‐year
design period was selected based on current ADOT new HMA pavement design practices.
Table 44 presents the key inputs for both the Materials Preliminary Engineering and Design
Manual (ADOT 1989) and the locally calibrated ADOT DARWin‐ME.
Figure 119. New JPCP Project Trial Design Structure
(TBD? = slab thickness to be determined during design)
Design Analysis
The researchers also used DARWin 3.1 software (software version of the 1993 AASHTO Guide for
Design of Pavement Structures) and DARWin‐ME in this design analysis, and modified both
145
software inputs and calibration coefficients as needed to be compatible with the Materials
Preliminary Engineering and Design Manual (ADOT 1989) and the locally calibrated pavement
models. Since this is a very heavily trafficked highway, both current Arizona and the DARWin‐ME
require dowels. The trial designs were run for 30 years. Results are presented in Table 45.
Figures 120 through 122 show predictions of transverse slab cracking, faulting, and IRI from the
local DARWin M‐E models.
The doweled JPCP designs produced from the Materials Preliminary Engineering and Design
Manual (ADOT 1989) show a 14.75‐inch PCC thickness. The locally calibrated DARWin‐ME for
Arizona produced a doweled 11‐inch PCC thickness. This considerably lower PCC thickness was
attributed to DARWin‐ME model’s ability to better optimize designs (material properties and
design features). The results presented show that the locally calibrated DARWin‐ME models and
design procedure do produce reasonable designs for Arizona if the right combination of material
inputs and design features is selected.
146
Table 44. Key Inputs for New JPCP Project Design
Input Category
Input Variables Inputs for Materials Preliminary Engineering and Design Manual
Inputs for Locally Calibrated ADOT DARWin‐ME
Structure and
materials
type and
properties
PCC
28‐day flexural strength: 672 psi*
28‐day elastic modulus:
4,286,826 psi*
28‐day flexural strength: 672 psi*
28‐day elastic modulus:
4,286,826 psi*
CTE: 4.5 × 10‐6 /° F
Aggregate base —
Aggregate base (AASHTO A‐1‐a)
Resilient modulus @ optimum
moisture content = 34,000 psi
Subgrade Modulus of subgrade reaction (k‐
value): 197 psi/inch
Subgrade type (AASHTO A‐2‐4)
Resilient modulus @ optimum
moisture content = 19,600 psi
Design
Joint spacing (ft) 13, 15, 17, 15 15
Dowel size (inch) 1.50 1.50
Slab width (ft) 12 12
Tied shoulder Yes Yes, 50% LTE
Design J factor 2.8 N/A
Design reliability 97% 97%
Drainage
coefficient 1.0 —
Traffic
Design period 30 years 30 years
Two‐way initial
AADTT 10,000 10,000
Directional
distribution 0.5 0.5
Lane distribution 0.8 0.8
Truck growth 3% compound 3% compound
Cumulative ESAL 127.8 million —
Cumulative trucks — 69.51 million
Performance
criteria
Initial
serviceability 4.2 —
Terminal
serviceability 3.0 —
Overall standard
deviation 0.25 —
Initial IRI (inch/mi) — 60
Terminal IRI
(inch/mi) — 150
Transverse
cracking (%) — 10
Joint faulting
(inch) — 0.12
*Computed based on a 28‐day compressive strength of 5000 psi.
147
Table 45. Trial Design Results for New JPCP Project Design
Key Variables Materials Preliminary Engineering
and Design Manual Locally Calibrated ADOT DARWin‐ME
Design ESALs 128 million 128 million
Cumulative trucks — 69.5 million
Dowel size Dowelled (J = 2.8) 1.25 inch
Reliability 97% 97%
Aggregate base
thickness 6 inch 6 inch
Design PCC thickness 14.75 inch 11 inch
Figure 120. Predicted Transverse Cracking for the New JPCP Trial Design Using DARWin‐ME
148
Figure 121. Predicted Faulting for the New JPCP Trial Design Using DARWin‐ME
Figure 122. Predicted IRI for the New JPCP Trial Design Using DARWin‐ME
149
CHAPTER 6. PRACTICAL METHODOLOGY TO RECALIBRATE DARWin‐ME
The local calibration documented in this report addresses key distresses and IRI associated with
HMA (and HMA overlays) and JPCP pavements in Arizona. Low temperature cracking of HMA
pavements could not be calibrated at this time, however, and should be further studied and
calibrated in the near future. Transverse fatigue cracking in ARFC/JPCP was validated using the
calibrated JPCP model. Two sections of CRCP and ARFC/CRCP were evaluated, and results
showed no reason not to use the global models at this time. In the future, ADOT may wish to
conduct further verifications, calibrations, and validations, including:
Reverification of the local Arizona distress and IRI models that were addressed under
this contract using additional updated performance data or other performance data that
becomes available.
Additional verification and calibration of ARFC/JPCP models for other distresses and IRI.
Calibration of other pavement structures including unbonded JPCP overlays, diamond
grinding projects, or HMA pavement full depth reclamation projects.
The AASHTO Guide for the Local Calibration of the Mechanistic‐Empirical Pavement Design
Guide (2010) provides a comprehensive methodology for recalibrating the DARWin‐ME models.
The guide presents a detailed, 11‐step roadmap for highway agencies to use for local calibration:
Step 1—Select hierarchical input levels.
Step 2—Develop experimental factorial design and matrix or sampling template.
Step 3—Estimate minimum sample size (number of pavement projects) required for
each distress/IRI prediction model validation and local calibration.
Steps 4—Select projects to populate sampling template.
Step 5—Extract and assess distress and project data.
Step 6—Conduct field and forensic investigations.
Step 7—Assess local bias: validation of global calibration values to local conditions,
policies, and materials.
Step 8—Eliminate local bias of distress and IRI prediction models.
Step 9—Assess the standard error of the estimate.
150
Step 10—Reduce standard error of the estimate.
Step 11—Interpret results and determine adequacy of ADOT MEPDG locally calibrated
models.
This process was adopted with modifications as needed for calibrating DARWin‐ME for local
Arizona conditions. The local calibration framework has been described throughout this report.
This chapter describes the guide’s key aspects as adapted as a framework for future model
recalibration or pavement calibration in Arizona. The framework presented may be used for
future recalibration of the DARWin‐ME models and design procedure for Arizona.
STEP 1: SELECT HIERARCHICAL INPUT LEVELS
The MEPDG allows for pavement design inputs (traffic, materials, climate, rehabilitation, etc.) to
be obtained at three hierarchical levels as described in Chapter 2.
The AASHTO Guide for the Local Calibration of the Mechanistic‐Empirical Pavement Design
Guide (2010) recommends selecting a hierarchical input level (1, 2, or 3) that is consistent with a
highway agency’s day‐to‐day practices for characterizing pavement design inputs and that is
sensitive to predicted performance and thus designs. Recommended hierarchical levels for
characterizing pavement materials properties, subgrade properties, and traffic for models
recalibration are presented in Tables 46 and 47.
Table 46. Recommended DARWin‐ME Hierarchical Input Levels for
New HMA and HMA Overlaid Pavements
MEPDG Input Variable Recommended Hierarchical
Input Level
HMA thickness Trial thickness
HMA CTE Levels 2 and 3
HMA dynamic modulus Level 3
HMA air voids in situ (at placement) Level 3
Effective HMA binder content Level 3
HMA creep compliance Level 1
HMA tensile strength Level 3
Base type/modulus Level 2
Base thickness Level 1
Subgrade type/modulus Level 1
Groundwater table Level 3
Climate Level 1
Truck volume Level 2
Truck axle load distribution Level 2
Tire load, contact area, and pressures Level 3
Truck speed Level 3
Truck wander Level 3
Initial IRI Level 2
151
Table 47. Recommended DARWin‐ME Hierarchical Input Levels for New JPCP
MEPDG Input Variable Recommended Hierarchical Input Level
PCC thickness Trial selected
PCC modulus of rupture and elasticity Levels 2 and 3
PCC CTE Levels 1 and 2
Joint spacing Level 1
Lane to PCC shoulder long‐term LTE Level 3
Edge support Level 1
Permanent curl/warp Level 3
Base type Level 1
Climate Level 1 Weather Station
Subgrade type/modulus Level 1
Truck axle load distribution Level 2
Truck volume Level 2
Tire pressure Level 3
Truck lateral offset Level 3
Truck wander Level 3
Initial IRI Level 2
STEPS 2 AND 3: DEVELOP EXPERIMENTAL DESIGN AND SAMPLING NEEDS
Experimental Design
Step 2 develops a statistical experiment to verify whether the DARWin‐ME global models and
procedure are adequate for Arizona local conditions and practices, and to calibrate the global
models if they are inadequate. To meet these objectives:
1. Define the problem: pavement types and models.
2. Define the population: project locations, materials, site, and design features.
3. Determine the sampling needs and extent: Use statistical procedures presented in the
AASHTO Guide for the Local Calibration of the Mechanistic‐Empirical Pavement Design
Guide (2010).
4. Establish the experimental design based on the outcome of Steps 1 through 3.
Examples of the experiment design sampling template and required sampling size are presented
in Tables 48 and 49.
152
Table 48. Sampling Template for New and Rehabilitated HMA Pavements
HMA Thickness (inches)
Granular Base Thickness (inches)*
Subgrade Type
Coarse‐Grained Soils (AASHTO Class A‐1
through A‐3)
Fine‐Grained Soils (AASHTO Class A‐4
through A‐7)
<8 <6 1 2
>6 3 4
>8 <6 5 6
>6 7 8
*All projects had granular base. Note: An equivalent dense‐graded aggregate base was assumed for permeable
asphalt‐treated bases (PATBs).
Table 49. Estimated Number of Pavement Projects Required for
Validation and Local Calibration
Pavement Type
Distress/IRI Distress/IRI Threshold (at 90% Reliability)
SEE
Minimum Projects
Required for Validation and Local Calibration
Minimum Projects Required
for Each Pavement Type
(n)*
New HMA and HMA overlaid HMA
Alligator cracking
20% lane area 5.01% 16
18 Transverse
thermal cracking
Crack spacing >100 ft of 630 ft/mi
150 ft/mi** 18
Rutting 0.4 inch 0.107 inches 14 IRI 169 inches/mi 18.9 inches/mi 80
New JPCP
Faulting <0.15 inch 0.033 inch 21
21 Transverse cracking
<10% slabs 4.52% 5
IRI 169 inches/mi 17.1 inches/mi 98
*2
2/
E
Zn
, where 2/Z = 1.601 (for a 90 percent confidence interval), = performance indicator
threshold (design criteria), and E = tolerable bias at 90 percent reliability (1.601*SEE).
**Estimated from other MEPDG implementation projects.
Note: In selecting the overall minimum number of required pavement HMA and JPCP projects, the
performance indicator IRI was excluded because the accuracy of the IRI models depends very much on the
accuracy of other pavement distress predictions. Sampling a vast number of projects to validate the IRI
models is therefore unnecessary if the individual distress prediction models are considered accurate and
reasonable.
153
STEPS 4 AND 5: SELECT HMA AND JPCP PAVEMENT PROJECTS TO POPULATE SAMPLING
TEMPLATE, AND EXTRACT AND EVALUATE DISTRESS AND PROJECT DATA
To populate the sampling templates developed in Step 3, projects are selected based on
recommendations from the AASHTO Guide for the Local Calibration of the Mechanistic‐Empirical
Pavement Design Guide (2010):
Projects should be representative of Arizona pavement design and construction
practices.
Projects should be representative of typical pavement condition (i.e., poor, moderate,
and good).
Project age should span the range typical of Arizona practice (i.e., newly constructed,
older existing, and rehabilitated).
Projects must be located throughout the state.
In general, information about local calibration projects is obtained from research experiments
and databases and state PMS databases. Criteria for final selection and inclusion into a project
database for recalibration of the DARWin‐ME models are availability of sufficiently detailed
input data (traffic, materials, and construction) with details comparable to the hierarchal levels
selected, and distress and IRI data in the formats required by the DARWin‐ME models (or
convertible to the formats without losing accuracy).
Once pavement selection is completed, pertinent project data are assembled in a project
database to be used for recalibration. This effort requires completing a detailed record review,
including a review of construction reports and design plans and reports; extracting historic
traffic volumes and vehicle class and axle load distributions from traffic databases; and
identifying and rectifying anomalies and errors in the assembled data.
STEP 6: CONDUCT FIELD AND FORENSIC INVESTIGATIONS
Projects that may be key to meeting the recalibration objective but lack all pertinent key data
(such as PCC strength and CTE) may require a forensic investigation to evaluate the given project
and collect additional data as needed to meet project objectives. Tasks performed as part of
forensic investigations include windshield distress survey, manual/automated distress survey,
nondestructive (FWD) deflection testing, and materials coring and lab texting/examination.
STEP 7: CHARACTERIZE BIAS AND GOODNESS OF FIT OF DARWin‐ME GLOBAL MODELS USED
UNDER ARIZONA CONDITIONS
Several methods (statistical or otherwise) may be used singly or in combination to verify the
MEPDG global models for Arizona local conditions. Statistical methods are used if there was a
154
good distribution of measured distress and IRI; when measured distress is mostly zero, a
nonstatistical approach to model validation is used, which consists of the following steps:
1. Execute the DARWin‐ME for each selected pavement project and predict pavement
distresses and IRI over a reasonable time period (typically 10 to 40 years).
2. Extract predicted distress and IRI data from the DARWin‐ME outputs that match age of
measured distress and IRI.
3. Perform statistical or nonstatistical analysis as appropriate to characterize bias and
prediction capacity.
4. Determine suitability of the DARWin‐ME global models for Arizona local conditions.
Details of the statistical or nonstatistical analysis as needed to verify model suitability are
presented in the following sections.
Evaluate Model Prediction Capacity (Goodness of Fit)
The goodness of fit of a given DARWin‐ME distress and IRI prediction model must be assessed by
determining R2 and SEE estimated using measured Arizona and DARWin‐ME predicted distress
and IRI data. Reasonableness of both diagnostic statistics is determined by comparing the
Arizona diagnostic statistics with those obtained for the DARWin‐ME global models using
Arizona data. Nearly all of the calibrated models showed significant improvement in prediction.
Engineering judgment must be used to determine the reasonableness of Arizona diagnostic
statistics.
Models exhibiting a poor R2 (i.e., R2 significantly less than the values presented in Table 50 for
the global models) or excessive SEE (significantly higher than the values presented in Table 50
for the global models) typically are considered inadequate for Arizona conditions.
Evaluate Model Bias
Bias is determined by performing linear regression using Arizona measured and DARWin‐ME
predicted distress and IRI, and performing the following two statistical hypothesis tests. A
significance level, , of 0.05 or 5 percent is typically assumed for all hypothesis testing.
Hypothesis 1: Paired t‐test. This test determines whether the Arizona measured and
DARWin‐ME predicted distress and IRI represents the same population:
1. Assume the following null and alternative hypothesis:
a. H0: mean measured distress/IRI = mean predicted distress/IRI.
b. HA: mean measured distress/IRI ≠ mean predicted distress/IRI.
155
Table 50. National Calibration under NCHRP 1‐40D of
New HMA Pavement and New JPCP Model Statistics
Pavement Type Performance
Model
Model Statistics
R2 SEE Number of Data
Points (N)
New HMA
Alligator cracking 0.275 5.01% 405
Transverse thermal
cracking
Level 1*: 0.344
Level 2*: 0.218
Level 3*: 0.057
— —
Rutting 0.58 0.107 inch 334
IRI 0.56 18.9 inches/mi 1926
New JPCP
Transverse slab
cracking 0.85 4.52% 1505
Transverse joint
faulting 0.58 0.033 inch 1239
IRI 0.60 17.1 inches/mi 163
*Level of inputs used for calibration.
2. Compute test p‐value. (This can be done using many types of statistical software
and spreadsheets such as MS Excel.)
3. Evaluate results by comparing the computed p‐value to the predetermined level of
significance (i.e., 0.05).
4. Interpret results: The null hypothesis H0 will be rejected if p‐value <0.05. Rejecting
H0 implies that Arizona measured and DARWin‐ME distress and IRI are essentially
from different populations at the 5 percent significance level, which indicates bias in
DARWin‐ME predicted distress and IRI because the measured distress and IRI are
considered to represent the ground truth for this test analysis.
Hypothesis 2: Slope of predicted and measured distress/IRI linear curve. This test
(which is only meaningful if hypothesis 1 was accepted) determines whether a linear
regression model (predicted distress/IRI = *measured distress/IRI) has a slope () of 1.0 at the 5 percent significance level. The test is as follows:
1. Using the results of the linear regression analysis, test the following null and
alternative hypotheses to determine if the linear regression model slope is 1.0:
a. H0: model slope () = 1.0.
b. HA: model slope () ≠ 1.0.
156
2. Compute test p‐value. (This can be done using many types of statistical software
and spreadsheets such as MS Excel.)
3. Evaluate results by comparing computed p‐value to the predetermined level of
significance for this test (i.e., 0.05).
4. Interpret results: The null hypothesis H0 is rejected if p‐value <0.05. Rejecting H0
implies that the linear model has a slope significantly different from 1.0 at the
5 percent significance level, which indicates that using the DARWin‐ME global
models outside of the range of Arizona measured distress and IRI will produce
biased predictions.
The presence of bias does not necessarily imply that the DARWin‐ME global models are
defective. It simply means that there is some bias in predicted distress and IRI values when the
DARWin‐ME is used under Arizona conditions.
Global DARWin‐ME models that are biased or have an inadequate prediction capacity or
goodness of fit need local calibration according to the procedure in Step 8.
STEP 8: PERFORM LOCAL CALIBRATION TO REDUCE BIAS AND IMPROVE GOODNESS OF FIT
Local calibration uses both linear and nonlinear regression procedures:
1. Split the assembled pavement projects data into two separate databases for calibration
and validation. Typically a 90/10 percent split is adopted. A 90/10 percent split implies
approximately 90 percent of all projects will be used to create a local calibration
database while the remaining 10 percent of projects will be used to create a validation
database. Projects must be selected at random while maintaining a general balance of
pavement type, design features, material properties, traffic, construction year,
location/climate, and so on. In essence the 10 percent validation database must
represent the same population as the 90 percent calibration database.
2. Perform a preliminary or tentative local calibration of the DARWin‐ME models using the
90 percent local calibration as follows:
a. Run the DARWin‐ME for all projects in the 90 percent local calibration database.
b. Assemble critical DARWin‐ME inputs (HMA thickness); DARWin‐ME outputs
(predicted damage, distress, and IRI); and measured distress and IRI required for
local calibration in an optimization database.
c. Determine DARWin‐ME models calibration coefficients that can be modified as
part of the local calibration. (See appendix.)
157
d. Perform optimization using linear and nonlinear regression techniques as
needed to select local calibration coefficients that maximize R2, minimize SEE,
and eliminates bias.
3. Validate the tentative locally calibrated MEPDG models as follows:
a. Run the DARWin‐ME for all projects in the 10 percent validation database using
the local calibration coefficients previously determined.
b. Assemble critical DARWin‐ME inputs (HMA thickness); DARWin‐ME outputs
(predicted damage, distress, and IRI); and measured distress and IRI required for
local calibration in an optimization database.
c. Determine goodness of fit and presence or absence of bias.
4. Modify the tentative locally calibrated DARWin‐ME models as needed until an adequate
model/solution is obtained (i.e., reasonable goodness of fit and no bias).
5. Develop final locally calibrated DARWin‐ME models using 100 percent of all projects.
Once local calibration is complete (DARWin‐ME’s local calibration coefficients are adjusted),
users must decide whether to develop new standard error models for use in reliability analysis.
In general if a small data set is used in local calibration, the global standard error models must
be left unchanged as there is not sufficient data to characterize models’ standard error.
However, with larger datasets, developing new standard error models is recommended
(guidance provided in the AASHTO Guide for the Local Calibration of the Mechanistic‐Empirical
Pavement Design Guide [2010]).
STEPS 9 THROUGH 11: ASSESS AND REDUCE STANDARD ERROR, AND INTERPRET RESULTS AND
DETERMINE ADEQUACY OF LOCALLY CALIBRATED DARWin‐ME MODELS
The final steps are to determine if the new local models meet engineering expectations by
performing a comprehensive sensitivity analysis and design comparisons with existing ADOT
pavement designs or in‐service pavements. The standard error provides an overall assessment
of the ability to predict the distress or IRI. It should be compared to the national or global
standard error. The local calibration standard error may be lower than the global value,
however, and that may be reasonable, but be careful to not accept too low of value that results
in an inadequate structural design.
158
Sensitivity Analysis
Sensitivity analysis involves the following steps:
1. Develop a baseline pavement design (a typical ADOT pavement design that reflects
current and future design and construction practices along with site conditions).
2. Identify key DARWin‐ME inputs (e.g., thickness, strength, traffic volume and climate).
3. Determine typical mean and likely range of key inputs. The typical mean values are used
in the baseline design.
4. Run the DARWin‐ME software for the baseline design.
5. Run the DARWin‐ME software changing each key input value individually within the
likely range of values.
6. Characterize change in distress and IRI values as the key input values change from the
mean.
7. Evaluate change in distress and IRI for reasonableness. (Does it meet engineering
expectations?)
If the change in distress and IRI meets engineering expectations, then the locally calibrated
models are considered reasonable. Otherwise, the models’ local coefficients may need to be
further adjusted to make predictions more reasonable.
Design Comparisons
Design comparisons involve using the new locally calibrated models within DARWin‐ME to
produce typical ADOT designs and determining if the new designs meet engineering
expectations. Existing ADOT designs or information from as‐constructed pavements in Arizona
can be used for these comparisons. Again, if the new DARWin‐ME designs do not meeting
engineering expectations (HMA or PCC sections are too thin or too thick), it may be necessary to
adjust the design reliability levels.
159
CHAPTER 7. ARIZONA INPUT DEFAULTS AND
CALIBRATION COEFFICIENTS FOR MEPDG
This chapter summarizes the information related to Arizona pavement site conditions, materials,
design features, and construction practices gathered for the ADOT DARWin‐ME implementation
effort. Data were gathered from various sources, including ADOT construction quality
assurance/quality control (QA/QC) databases and LTPP.
DEFAULT LEVEL 3 INPUTS
HMA Materials
Default Level 3 inputs (HMA gradation, in‐situ air voids, effective volumetric binder content, and
unit weight) for HMA materials were developed from ADOT’s QA/QC databases. These inputs
were computed from more than 300 lab and field measured data, including distributions, mean,
ranges, and standard deviation. The computed defaults are presented in Table 51.
PCC Materials
Default Level 3 inputs (compressive strength, flexural strength, elastic modulus, CTE, and
mixture constituents) for PCC materials inputs were developed from LTPP materials databases.
The PCC defaults were computed using lab measured data from the LTPP GPS experiments,
which is more representative of ADOT practices. A summary of the computed defaults are
presented in Tables 52 through 56.
Table 51. Default Level 3 Inputs for HMA Materials
Material ADOT
Material Designation
Gradation (%) In situ Air
Voids (%)
Unit Weight (pcf)
Effective Binder Content (%)
Passing ¾‐inch Sieve
Passing ⅜‐inch Sieve
Passing No. 4 Sieve
Passing No. 200 Sieve
1/2‐inch HMA mix
M 12 100.0 79.2 51.3 4.3 7.7 139.8 11.9
M 12F 100.0 82.2 59.8 4.1 7.7 138.7 12.6
M 12K 100.0 78.9 49.5 4.6 7.3 141.9 11.1
3/4‐inch HMA mix
M 34 97.6 73.1 55.8 4.4 7.6 141.9 10.8
M 34A 98.3 72.7 54.2 4.3 7.0 142.0 11.3
M 34B 99.8 77.3 60.8 4.9 7.5 138.1 10.7
M 34F 98.9 73.7 58.7 4.3 6.9 142.0 11.4
M 34K 99.1 67.4 43.0 4.7 6.8 144.8 10.0
M 34KA 99.2 69.4 42.0 4.3 7.0 146.3 8.9
M 34KB 98.0 67.8 43.8 4.5 6.7 154.5 4.6
ATB mix M BM 92.9 67.4 51.5 3.8 6.9 143.1 10.9
Road mix M RD 100.0 71.5 36.3 2.5 7.9 134.0 15.0
ACFC ACFC 100 99 41.1 1.48 — — —
AR‐ACFC AR‐ACFC 100 99 37.7 1.4 — — —
RF RF 100 99 35 0.9 — — —
160
Table 52. Default Level 3 PCC Compressive Inputs
SHRP ID Test No. Compressive Strength
(Long Term) (psi)
Compressive Strength
(28 days) (psi*)
0601 1 5,540 3,847
0601 2 7,040 4,889
0603 1 6,070 4,215
0603 2 6,800 4,722
0604 1 6,170 4,285
0604 2 8,310 5,771
0605 2 5,440 3,778
0607 2 5,810 4,035
0608 1 6,620 4,597
0608 2 6,930 4,813
0659 2 6,800 4,722
0660 1 6,020 4,181
0662 1 5,450 3,785
0662 2 8,210 5,701
7079 1 6,460 4,486
7079 2 7,110 4,938
7614 1 6,020 4,181
7614 2 5,090 3,535
Average 6,438 4,471
*Converted using DARWin‐ME strength growth relationship.
Table 53. Default Level 3 PCC Flexural Strength Inputs
Project ID Route ID
Compressive
Strength
(28 days) (psi)
Unit Weight
(lb/yd3)
Water/Cement
Ratio
Cementitious
Materials (lb)
H087502C L202 5090 147 0.47 581
H323002C L101 4900 149 0.46 586
H406001C L101 5307 152 0.47 581
H458401C I‐40 4664 148 0.49 595
H473301C L101 5280 148 0.48 588
H484501C L101 5217 N/A N/A N/A
H538101C L202 5707 150 0.36 586
H575601C L101 5513 149 0.47 581
H591501C L202 4988 150 0.40 581
H601401C I‐10 5383 151 0.40 581
Mean 5205 149 0.44 584
Note: These are current PCC strength values since incentives were introduced.
161
Table 54. Default Level 3 PCC Elastic Modulus Inputs
SHRP ID Test No. Poisson Ratio Unit Weight
(pcf)
Elastic Modulus
(Long Term) (psi)
Elastic Modulus
(28‐day) (psi)
7079 1 0.21 145 4,150,000 3,458,333
7079 2 0.10 149 3,750,000 3,125,000
7613 1 0.26 145 3,400,000 2,833,333
7613 2 0.12 149 4,850,000 4,041,667
7614 1 0.13 147 4,000,000 3,333,333
7614 2 0.10 149 4,900,000 4,083,333
Average 0.15 147 4,175,000 3,479,167
Table 55. Default Level 3 PCC CTE Inputs
SHRP ID CTE (o F/o F/inch) Coarse
Aggregate Type Mean CTE (o F/o F/inch)
0213 4.8 Granite
4.78
0214 4.8 Granite
0215 4.5 Granite
0216 4.42 Granite
0217 4.8 Granite
0218 4.8 Granite
0219 4.8 Granite
0220 4.8 Granite
7613 4.63 Granite
7614 5.5 Granite
0221 4.4 Limestone
4.4 0222 4.4 Limestone
0223 4.4 Limestone
0224 4.4 Limestone
Note: These are the corrected CTE values obtained from the FHWA laboratory.
Table 56. Default Level 3 PCC Mixture Constituents Inputs
SHRP
ID
Coarse
Content
(lb/yd3)
Fine
Content
(lb/yd3)
Cement
Content
(lb/yd3)
Water Content
(lb/yd3)
Water‐to‐
Cement
Ratio
7079 1311 1809 706 283 0.40
7614 1858 1265 513 240 0.47
162
Chemically Treated Aggregate (Cement Aggregate Mixture) Base Materials
Default Level 3 inputs (elastic modulus) for key cement aggregate mixture materials (used as
cement‐treated bases) were developed from only a few pavement projects available in the LTPP
materials database. These computed defaults are presented in Table 57.
Table 57. Default Level 3 Inputs for Cement Aggregate Mixture Materials
SHRP ID Test No.
Compressive Strength
(Long Term) (psi)
Elastic Modulus
(Long Term) (psi)
0604 2 720 1,529,470
0605 2 610 1,407,796
0607 2 790 1,602,096
0608 1 810 1,622,248
0608 2 920 1,728,896
7614 1 2,370 2,774,911
7614 2 2,550 2,878,359
Average 1,253 1,934,825
Unbound Aggregate Base Materials
Default Level 3 inputs (gradation and Atterberg limits) for unbound aggregate base and subbase
materials were developed from more than 300 lab and field measured data in ADOT’s QA/QC
databases. These inputs are summarized in Table 58.
163
Table 58. Default Level 3 Inputs for Unbound Aggregate Materials
Material Property ADOT Material Code
Aggregate Base Aggregate Subbase Embankment
Plasticity index 2.5 8.4 8.8
Plastic limit 5.4 19.6 14.6
Liquid limit 7.9 28.1 23.4
Passing No. 200 6.3 6.9 32.2
Passing No. 100 9.3 9.6 41.5
Passing No. 50 15 14.2 50.3
Passing No. 40 19.1 17.4 54.4
Passing No. 30 24.4 21.5 58.9
Passing No. 16 32.9 28.4 65.7
Passing No. 10 39.7 33.9 70.9
Passing No. 8 42.1 35.7 72.6
Passing No. 4 52.5 44.8 78.7
Passing 3/8 68.2 59.9 89.3
Passing 1/2 76.8 67.3 91.3
Passing 3/4 92.9 81.2 94
Passing 1 99.5 90.4 95.6
Passing 1‐1/2 99.9 98.4 97.4
Passing 2 99.97 99.8 98.3
Passing 2‐1/2 99.98 99.98 98.9
Passing 3 99.99 100 99.3
Passing 3‐1/2 100 100 100
Subgrade Soil Materials
Default Level 3 inputs (gradation, Atterberg limits, and resilient modulus at optimum moisture
content) for subgrade materials were developed from DARWin‐ME, LTPP, and ADOT projects.
Default gradations and Atterberg limits were obtained from DARWin‐ME (global defaults).
Resilient modulus at optimum moisture content was obtained through backcalculation from
nearly all Arizona sections. The researchers used field FWD tests to backcalculate the in situ
subgrade elastic modulus for each flexible pavement, which was then adjusted to obtain
estimates of Mr at optimum moisture content (Table 59). Field FWD tests were also used to
backcalculate the in situ subgrade dynamic k‐value for each rigid pavement, which was then
used to obtain the Mr at optimum moisture content through iteration of the DARWin‐ME to
match the output k‐value with the field backcalculated k‐value. Table 59 summarizes the mean
Mr values at optimum moisture content for each AASHTO soil classification.
164
Table 59. Default Level 3 Inputs for Subgrade Materials Mr at Optimum Moisture Content
AASHTO Soil
Classification
Resilient Modulus at Optimum Moisture (psi)
Base/Subbase for Flexible
and Rigid Pavements
Embankment and
Subgrade for Flexible
Pavements
Embankment and
Subgrade for Rigid
Pavements
A‐1‐a 34,000 18,700 18,700
A‐1‐b 34,000 18,700 18,700
A‐2‐4 N/A 15,000 19,600
A‐2‐5 N/A 15,300 15,900
A‐2‐6 N/A 15,800 15,000
A‐2‐7 N/A 16,500 16,000
A‐3 N/A 15,000 15,000
A‐4 N/A 14,400 16,500
A‐5 N/A 14,500 14,500
A‐6 N/A 14,900 14,600
A‐7‐5 N/A 15,000 15,000
A‐7‐6 N/A 14,800 16,300
Note: These resilient modulus values were obtained from averaging over all sections for flexible and rigid pavements
used in the Arizona calibration database. The FWD backcalculated in situ elastic solid subgrade modulus for flexible
pavement was adjusted to optimum moisture and a lab value by multiplying by 0.55 for fine‐grained soils and 0.67 for
coarse‐grained soils as done for the national calibration. The FWD backcalculated in situ k‐value for rigid pavement
was used to establish through iteration the proper input resilient modulus at optimum moisture. Some soils were
missing in the testing and were assigned values.
Traffic
Default traffic inputs for DARWin‐ME were developed under the ADOT project, Development of
a Traffic Data Input System in Arizona for the MEPDG (Darter et al. 2010). The project objectives
were to:
Identify the needs of various sections within ADOT in terms of traffic data specifically
related to the 1993 AASHTO Guide for Design of Pavement Structures and the MEPDG.
Evaluate the current ADOT practice in terms of obtaining, compiling, and managing
traffic data.
Critically investigate the existing traffic data collection infrastructures, such as weigh‐in‐
motion (WIM) stations, and determine their validity and usefulness for the MEPDG.
Develop a detailed action plan for ADOT to continuously obtain all necessary traffic data
and compile that information for effective use in the MEPDG. The action plan should
also include a detailed cost estimate.
165
The default vehicle class distributions, axle load distributions, and other key traffic inputs from
the ADOT project, Development of a Traffic Data Input System in Arizona for the MEPDG (Darter
et al. 2010), were used to locally calibrate the DARWin‐ME models. The project’s final report
contains all the information required to obtain default traffic inputs.
Climate
Default climate inputs were obtained from NCDC climate data from several Arizona weather
stations. Many weather stations were missing some of the hourly data. Each weather station
was checked and fixed if deficient. The default climate data is included in DARWin‐ME by
default.
DARWin‐ME ARIZONA LOCAL CALIBRATION COEFFICIENTS
The researchers conducted an extensive local calibration and validation effort to ensure that the
distress and IRI models were unbiased (i.e., did not, on average, over‐ or underpredict rutting,
fatigue cracking, and IRI). The Arizona specific local calibration coefficients have been entered
into the most current version of the DARWin‐ME software, however, Arizona designers will need
to update the local calibration coefficients with each new version that AASHTO provides. The
coefficients used are always output with every run of the software and located under the
Calibration tab in the Excel output and at the end of the PDF output file.
Tables 60 through 62 summarize the 2012‐derived local calibration coefficients for the
HMA/HMA overlays and JPCP. Composite pavements should follow the JPCP local calibration
coefficients for concrete and HMA local calibration coefficients for HMA pavements. There was
insufficient data to calibrate CRCP models, but there is no reason to believe the global
coefficients will not be adequate in Arizona. The Arizona local calibration coefficients may be
changed in the future as ADOT conducts additional local calibration efforts.
166
Table 60. DARWin‐ME Local Calibration Coefficients for New HMA and HMA Overlaid HMA Pavement
Model or Submodel Type
Model Coefficients ADOT Local Calibration Change from Global Models
Fatigue damage model (AC fatigue)
K1 0.007566 No
K2 3.9492 No
K3 1.281 No
BF1 249.0087232 Yes
BF2 1 No
BF3 1.233411397 Yes
Alligator cracking model (AC cracking bottom)
C1 1 No
C2 4.5 Yes
C4 6000 No
AC cracking bottom standard deviation
1.1+22.9/(1+exp(‐0.1214‐2.0565*LOG10(BOTTOM+0.0001)))
Yes
Reflection cracking model
a 3.5+0.75heff No
b ‐0.688584‐3.37302*heff‐0.915469 No
c 2.55 Yes
d 1.23 Yes
AC rutting
K1 ‐3.35412 No
K2 1.5606 No
K3 0.4791 No
BR1 0.69 Yes
BR2 1 No
BR3 1 No
AC rutting standard deviation
0.0999*Pow(RUT,0.174)+0.001 Yes
Base rutting (granular subgrade
rutting)
K1 (base) 2.03 No
BS1 (base) 0.14 Yes
Base rutting standard deviation
0.05*Pow(BASERUT,0.115)+0.001 Yes
Subgrade rutting (fine subgrade rutting)
K1 (subgrade) 1.35 No
BS1 (subgrade) 0.37 Yes
Subgrade rutting standard deviation
0.05*Pow(SUBRUT,0.085)+0.001 Yes
HMA transverse cracking model (thermal fracture)
Thermal fracture level 1K
1.5 No
Thermal fracture level 1 standard deviation
0.1468 * THERMAL + 65.027 No
Thermal fracture level 2K
0.5 No
Thermal fracture level 2 standard deviation
0.2841 *THERMAL + 55.462 No
Thermal fracture level 3K
1.5 No
Thermal fracture level 3 standard deviation
0.3972 * THERMAL + 20.422 No
HMA IRI model
C1 (for rutting) 1.2281 Yes
C2 (for fatigue) 0.1175 Yes
C3 (for transverse) 0.008 No
C4 (for SF) 0.0280 Yes
167
Table 61. DARWin‐ME Local Calibration Coefficients for New JPCP
Model or
Submodel Type Model Coefficients ADOT Local Calibration
Change from
NCHRP 1‐40D
Global Models
PCC fatigue model C1 2 No
C2 1.22 No
JPCP transverse
cracking model
C4 0.19 Yes
C5 ‐2.067 Yes
PCC cracking
standard deviation Pow(9.87*CRACK,0.4012)+0.5 Yes
Faulting model
C1 0.0355 Yes
C2 0.1147 Yes
C3 0.00436 Yes
C4 1.1E‐07 Yes
C5 20000 Yes
C6 2.0389 Yes
C7 0.1890 Yes
C8 400 No
PCC faulting
standard deviation Pow(0.037*FAULT,0.6532)+0.001 Yes
JPCP IRI model
J1 (for cracking) 0.60 Yes
J2 (for spalling) 3.48 Yes
J3 (for faulting) 1.22 Yes
J4 (for SF) 45.20 Yes
PCC IRI JPCP
standard deviation 5.4 No
Table 62. DARWin‐ME Local Calibration Coefficients for New CRCP
Model or Submodel Type
Model Coefficients ADOT Local Calibration
Change from
NCHRP 1‐40D
Global Models
PCC fatigue model C1 2 No
C2 1.22 No
CRCP punchout model
C3 85 Yes
C4 1.4149 Yes
C5 ‐0.8061 Yes
PCC punchout standard deviation
1.5+2.9622*Pow(PO,0.4356) Yes
CRCP IRI model
C1 3.15 No
C2 28.35 No
PCC IRI CRCP standard deviation
5.4 No
168
CHECKLIST OF ALL INPUTS
Below are summaries of key data inputs for flexible (new HMA and HMA overlays of existing
HMA) pavements (Table 63) and new JPCP (Table 64).
Table 63. Checklist of Flexible Pavement Inputs
Input Parameter ADOT User’s Guide (2012)
Comment
General information
Design life Table 1, page 13
Base/subgrade construction data
Table 2, page 14
Pavement construction date
Traffic opening date
Performance criteria
Initial IRI Table 7, page 19
Reliability level Table 8, page 22
Terminal IRI
Table 6, page 18
Top‐down fatigue cracking Not used
Bottom‐up fatigue cracking
Transverse cracks
Total rut depth
HMA rut depth
Reflective cracking Regression
Traffic, site features
Average annual daily truck traffic
Table 9, page 27
Number of lanes in design direction
Percent of two‐directional trucks in design direction
Percent of trucks in design lane
Operational speed
General traffic, axle configuration
Average axle width
Table 26, page 49
Default value
Dual tire spacing Default value
Dual tire pressure
Tandem axle spacing Default value
Tridem axle spacing Default value
quad axle spacing Default value
Traffic, lateral wander
Mean wheel location
Page 49
N/A
Wander, standard deviation Default value
Design lane width N/A
Traffic, wheelbase Average spacing
Table 27, page 49 N/A Percent trucks
Traffic, volume
Normalized vehicle class distribution Table 11, page 30
Growth rate and function Page 32
Monthly adjustment factors Table 10, page 28
Number of axles per truck type Table 25, page 48 Default value
Hourly distribution factors Table 12, page 31 N/A
Single axles
Tables 13 to 24, pages 36 to 47
Tandem axles
Tridem axles
Quad axles
169
Table 63. Checklist of Flexible Pavement Inputs (Continued)
Input Parameter ADOT User’s Guide (2012)
Comment
Climate
Weather station Figure 18, page 55
Location: latitude, longitude, elevation
Depth to water table Figure 19, page 56
HMA layers
Summary Table 31, page 58
Thickness
Dynamic modulus; Level 1 Table 35, page 64
Creep compliance; Level 1 Page 67
Tensile strength; Level 1 Page 66
Asphalt binder grade Page 65
Gradation Table 36, page 65
Effective asphalt content by volume
Table 37, page 66
Air voids
Unit weight
Poisson’s ratio Table 38, page 67 Default Value
Surface shortwave absorptivity Page 67 Default Value
Thermal conductivity Page 68 Default Value
Heat capacity Page 68 Default Value
Coefficient of thermal contraction Page 68 Default Value
Existing HMA – backcalculated modulus
Unbound
aggregate and soil
layers
Summary Table 34, page 61
Thickness
AASHTO soil classification
Resilient modulus Table 50, page 77 Table 51, page 82
Poisson’s ratio Table 52, page 83 Table 53, page 84
Default Value
Coefficient of lateral pressure Default Value
Specific gravity Page 82 Default Value
Saturated hydraulic conductivity Page 82 Default Value
Soil‐water characteristic curve Page 82 Default Value
Water content; optimum Page 82
Dry unit weight; modified proctor Page 82
Gradation Table 52, page 83 Table 53, page 84
Plasticity index
Liquid limit
Local calibration factors
Alligator cracking; bottom‐up cracking Table 62, page 110
HMA rutting Table 63, page 111
Aggregate base rutting; coarse subgrade Table 63, page 111
Subgrade rutting; fine subgrade Table 63, page 111
HMA IRI regression equation Table 64, page 112
Reflection cracking Table 62, page 110
170
Table 64. Checklist of Rigid Pavement Inputs
Input Parameter ADOT User’s Guide
(2012) Comment
General information
Design life Table 1, page 13
Pavement PCC construction date
Traffic opening date
Performance criteria
Initial IRI Table 7, page 19
Reliability Level Table 8, page 22
Terminal IRI
Table 6, page 18
JPCP transverse fatigue cracking
JPCP transverse joint faulting
JPCP transverse joint load transfer
CRCP punchouts
CRCP crack width 50% Rel
CRCP crack load transfer 50% Rel
Traffic, site features
AADTT
Table 9, page 27
Number of lanes in design direction
Percent of two‐directional trucks in design direction
Percent of trucks in design lane
Operational speed
General traffic, axle configuration
Average axle width
Table 26, page 49
Default Value
Dual tire spacing Default Value
Dual tire pressure
Tandem axle spacing Default Value
Tridem axle spacing Default Value
Quad axle spacing Default Value
Traffic, lateral
wander
Mean wheel location
Page 49
NA
Wander, standard deviation Default Value
Design lane width NA
Traffic, wheelbase Average spacing
Table 27, page 49 NA Percent trucks
Traffic, volume
Normalized vehicle class distribution Table 11, page 30
Growth rate and function Page 32
Monthly adjustment factors Table 10, page 28
Number of axles per truck type Table 25, page 48 Default Value
Hourly adjustment factors Table 12, page 31 NA
Single axles
Tables 13 to 24, pages 36 to 47
Tandem axles
Tridem axles
Quad axles
171
Table 64. Checklist of Rigid Pavement Inputs (Continued)
Input Parameter ADOT User’s Guide
(2012) Comment
Climate
Weather station Figure 18, page 55
Location: latitude, longitude Long‐Lat.com
Elevation of project Long‐Lat.com
Depth to water table Figure 19, page 56
HMA layers
Summary Table 31, page 58
Thickness
Dynamic modulus; Level 1 Table 35, page 64
Creep compliance; Level 1 Page 67
Tensile strength; Level 1 Page 66
Asphalt binder grade Page 65
Gradation Table 36, page 65
Effective asphalt content by volume
Table 37, page 66
Air voids
Unit weight
Poisson’s ratio Table 38, page 67 Default value
Surface shortwave absorptivity Page 67 Default value
Thermal conductivity Page 68 Default value
Heat capacity Page 68 Default value
Coefficient of thermal contraction Page 68 Default value
Existing HMA – backcalculated modulus
Unbound aggregate and soil layers
Summary Table 34, page 61
Thickness
AASHTO soil classification
Resilient modulus Table 50, page 77 Table 51, page 82
Poisson’s ratio Table 52, page 83 Table 53, page 84
Default value
Coefficient of lateral pressure Default value
Specific gravity Page 82 Default value
Saturated hydraulic conductivity Page 82 Default value
Soil‐water characteristic curve Page 82 Default value
Water content, optimum Page 82
Dry unit weight; modified proctor Page 82
Gradation Table 52, page 83 Table 53, page 84
Plasticity index
Liquid limit
Local calibration factors
JPCP IRI
JPCP transverse fatigue cracking
JPCP transverse joint faulting
CRCP IRI
CRCP punchouts
CRCP crack width
172
173
CHAPTER 8. SUMMARY AND RECOMMENDATIONS
The complete implementation process has involved several major efforts to assure ADOT that
the MEPDG models will predict distress and IRI that matches Arizona experience:
Develop a user’s guide with procedures for creating proper inputs that use the MEPDG
to design new, reconstructed, and rehabilitated pavement structures.
Use Arizona performance data to verify, validate, and calibrate the MEPDG models (if
necessary) to remove bias (consistent over‐ or underprediction) and improve accuracy
of prediction. This was accomplished through the verification, validation, and
recalibration of Arizona calibration coefficients. In nearly all cases, this effort improved
the accuracy of the distress and IRI models and removed the over‐ and underprediction
of bias.
Develop recommended levels of design reliability and model standard deviations
specific to each distress and IRI model that are required to design a pavement with a
desired level of reliability.
Conduct design comparisons and sensitivity studies that help to establish confidence in
the pavement design results achieved by the MEPDG.
Summary
The ADOT User’s Guide (2012) presents the following information to help ADOT pavement
design engineers and others use the AASHTO DARWin‐ME pavement design guide and software:
An overview of the AASHTO DARWin‐ME procedure.
Help information for installing the software.
Guidelines for obtaining all needed inputs.
Design guidance for using the software with the following pavement types:
o New or reconstructed HMA pavement.
o New or reconstructed JPCP and CRCP.
o New or reconstructed composite pavement (ARFC overlay on JPCP or CRCP).
o HMA rehabilitation: HMA overlay on existing HMA.
o HMA rehabilitation: HMA on existing JPCP.
o JPCP rehabilitation: CPR diamond grinding and JPCP overlay on an existing HMA or
JPCP pavement.
174
Examples of pavement design using the design guide software.
o New or reconstructed HMA pavement.
o New or reconstructed JPCP.
o New or reconstructed ARFC/JPCP.
o HMA overlay on existing HMA.
The MEPDG prediction models were verified, validated, and, if necessary, recalibrated using
Arizona pavement sections. One hundred and eighty HMA, JPCP, ARFC/JPCP, and rehabilitated
pavements were included in a valuable database that represents Arizona pavement
performance over many years. This database was used to verify, validate, and recalibrate the
prediction models to make them more accurate and unbiased (neither over‐ nor
underprediction). They were also used to establish Arizona design inputs and the appropriate
standard deviation or error of each model for use in reliability design, which will make it possible
to design a pavement in Arizona with the desired reliability at the most optimum cost possible.
Table 65 shows the global model goodness of fit statistics and the Arizona specific calibration
goodness of fit statistics. The improvement in these models through verification, validation, and
calibration with Arizona data is well‐documented in this table; the calibrated models will provide
much more accurate, reliable, and cost‐effective designs than the global calibrations.
175
Table 65. Comparison of MEPDG Global and Arizona‐Specific
Model Goodness of Fit Statistics
Pavement Type
Distress/IRI Models
Global Models ADOT Calibrated Models
Global R2
(%)* Global Model
SEE* Arizona R2
(%) Arizona SEE
New HMA and HMA Overlay
Alligator cracking
8 14% 58 13%
Transverse thermal cracking
N/A N/A Not calibrated Not calibrated
Total rutting 5 0.31 inch 21 0.12 inch
IRI 30 19 inches/mi 80 8 inches/mi
New JPCP
Transverse cracking
20 9% 78 6%
Transverse joint faulting
45 0.03 inch 52 0.03 inch
IRI 35 25 inches/mi 81 10 inches/mi
ARFC/JPCP
Transverse cracking
Same as JPCP Same as JPCP Same as JPCP Same as JPCP
Rutting Same as HMA Same as HMA
Same as ADOT HMA
Same as ADOTHMA
IRI Same as HMA Same as HMA
Same as ADOT HMA
Same as ADOTHMA
New CRCP
Punchouts 68% 5 PO/mi Same as global
2 ADOT CRCP matched global predictions
Slab/base friction
Established values
N/A Same as global N/A
*Global calibration coefficients using the Arizona database.
Information presented in Table 35 shows that the local models also have improved goodness of
fit statistics. The results of the design comparisons also show reasonable distress and IRI
predictions by the Arizona locally calibrated MEPDG.
The Arizona calibrated models had a lower model prediction error, improved accuracy and
reliability, and reasonable sensitivity to changes in inputs. The locally calibrated MEPDG models
for Arizona are considered reasonable and are recommended for use in pavement design.
176
Recommendations
The following improvements are recommended to further improve pavement design using
AASHTO DARWin‐ME:
Evaluation of Arizona State University (ASU) lab testing data for binder and HMA
mixtures for use as defaults to improve estimation of HMA dynamic modulus. ASU
conducted substantial testing under the ADOT project, Development of Performance
Related Specifications for Asphalt Pavements in the State of Arizona (Witczak 2008), that
may better characterize HMA mixes.
Further evaluation of the low‐temperature cracking model and its calibration in
Arizona. This model was not calibrated under this study due to lack of creep compliance
and indirect tensile strength data for ADOT HMA mixes. In addition, there was
insufficient information to properly characterize observed HMA transverse cracking
distress (such as thermal fracture and shrinkage). However in Colorado and Wyoming,
DARWin‐ME implementation efforts produced a local calibration coefficient K value of
7.5. This may be a reasonable value for ADOT to use while additional research is
conducted to more fully establish proper local calibration coefficients.
Research into the cause of transverse cracking of HMA pavements in desert warm
nonfreezing areas. There is substantial transverse cracking in nonfreeze areas that is not
considered in the DARWin‐ME. These cracks lead to roughness and increased IRI that is
not considered in design. The mechanism may be shrinkage of the HMA mixture due in
part to absorption of binder into the porous aggregates.
Major increase in the collection and analysis of truck traffic data. Sufficient current
WIM axle weight data and truck classification data are not available. This will require
additional WIM sites across the state.
Consideration of AR mixtures. These materials were not included in the DARWin‐ME
development. Since Arizona uses substantial quantities of AR mixes in new construction
and overlays, better characterization would be useful in design.
177
REFERENCES
American Association of State Highway and Transportation Officials (AASHTO). 1993. Guide for
Design of Pavement Structures. Washington, D.C.: American Association of State Highway
and Transportation Officials.
American Association of State Highway and Transportation Officials (AASHTO). 2008.
Mechanistic‐Empirical Pavement Design Guide, A Manual of Practice, July 2008, Interim
Edition. Washington, D.C.: American Association of State Highway and Transportation
Officials.
American Association of State Highway and Transportation Officials (AASHTO). November 2010.
Guide for the Local Calibration of the Mechanistic‐Empirical Pavement Design Guide.
Washington, D.C.: American Association of State Highway and Transportation Officials.
Arizona Department of Transportation (ADOT). March 1989. Materials Preliminary Engineering
and Design Manual. Phoenix: Arizona Department of Transportation.
Darter, M. I., L. Titus‐Glover, and D. J. Wolf. 2010. Development of a Traffic Data Input System in
Arizona for the MEPDG. Publication FHWA‐AZ‐13‐672. Phoenix: Arizona Department of
Transportation.
National Cooperative Highway Research Program (NCHRP). 2004. 2002 Design Guide: Design of
New and Rehabilitated Pavement Structures. Project 1‐37A. Publication FHWA‐RD‐00‐129.
Washington, D.C.: National Cooperative Highway Research Program.
http://onlinepubs.trb.org/onlinepubs/archive/mepdg/guide.htm
National Cooperative Highway Research Program (NCHRP). 2007. Version 1.0 – Mechanistic‐
Empirical Pavement Design Guide Software. Washington, D.C.: National Cooperative
Highway Research Program.
http://onlinepubs.trb.org/onlinepubs/nchrp/nchrp_rrd_308.pdf
Smith, K. D., D. G. Peshkin, A. L. Mueller, E. Owusu‐Antwi, and M. I. Darter. 1991. Evaluation of
Concrete Pavements in the Phoenix Urban Corridor. Publication FHWA‐AZ91‐264‐1. Phoenix:
Arizona Department of Transportation.
Witczak, M. W. 2008. Development of Performance Related Specifications for Asphalt Pavements
in the State of Arizona. Publication FHWA‐SPR‐08‐402‐2. Phoenix: Arizona Department of
Transportation.
178
179
APPENDIX: DARWin‐ME HMA PAVEMENT AND
JPCP PERFORMANCE PREDICTION MODELS
A brief description of the Mechanistic‐Empirical Pavement Design Guide (MEPDG) models used
to predict performance is presented in this appendix. Additional information is available in the
original MEPDG documentation (NCHRP 2004, 2007; AASHTO 2008).
NEW AND RECONSTRUCTED HMA PAVEMENTS
Alligator Cracking
Alligator cracking prediction in the MEPDG comprises several submodels and algorithms for
estimating critical horizontal strain at the bottom of the HMA layers, empirical models relating
strain to allowable number of load repetitions, and transfer functions for computation
incrementally of HMA bottom‐up fatigue damage by dividing the actual number of axle loads by
the allowable number of axle loads, and an alligator cracking model based on estimated fatigue
damage levels. Eq. 15 through Eq. 22 present the key models used by the MEPDG for predicting
alligator cracking (AASHTO 2008).
Fatigue Damage Model
This model computes accumulated fatigue damage DI in the HMA over time.
TplmjHMAfTplmj N
nDIDI
,,,,
,,,,
(Eq. 15)
Where
N = actual number of axle load applications within a specific time period
j = axle load interval
m = axle load type (single, tandem, tridem, quad, or special axle configuration)
l = truck type using the truck classification groups included in the MEPDG
p = month
T = median temperature for the five temperature intervals or quintiles used to
subdivide each month (° F)
Nf‐HMA = allowable number of axle load applications for a flexible pavement and
HMA overlays to fatigue cracking
180
Allowable Number of Axle Load Applications Model
3322
11ffff k
HMAk
tfHfHMAf ECCkN (Eq. 16)
Where
Nf‐HMA = allowable number of axle load applications for a flexible pavement and HMA
overlays to fatigue cracking
εt = tensile strain at critical locations and calculated by the
structural response model (inch/inch)
EHMA = dynamic modulus of the HMA measured in compression (psi)
kf1, kf2, kf3 = global field calibration parameters (kf1 = 0.007566, kf2 = ‐3.9492,
and kf3 = ‐1.281)
βf1, βf2, βf3 = local or mixture‐specific field calibration constants; for the global
calibration effort, these constants were set to 1.0
MC 10 (Eq. 17)
69.084.4
bea
be
VV
VM (Eq. 18)
Where
Vbe = effective asphalt content by volume (percent)
Va = percent air voids in the HMA mixture (in situ only, not mixture design)
CH = thickness correction term as follows:
HMAH
H
e
C
49.302.111
003602.0000398.0
1
(Eq. 19)
Where
HHMA = total HMA thickness (inch)
181
For the HMA alligator cracking allowable number of axle load applications model, the MEPDG
allows the following model coefficients to be modified as needed as part of Arizona local
calibration to improve alligator cracking prediction accuracy and remove bias:
kf1, kf2, kf3 = global field calibration parameters (kf1 = 0.007566, kf2 = ‐3.9492, and kf3
‐1.281)
βf1, βf2, βf3 = Arizona local or mixture specific field calibration constants were set to 249.0087,
1.0, and 1.2334
Alligator Cracking Model
This S‐shaped model predicts alligator fatigue bottom‐up cracking FCBottom in the traffic lane due
to accumulated fatigue damage DIBottom.
))((
4*22
*11160
1BottomDILogCCCCBottom
e
CFC (Eq. 20)
Where
FCBottom = Area of alligator cracking that initiates at the bottom of the
HMA layers (percent of total lane area)
DIBottom = Cumulative damage index at the bottom of the HMA layer
C1,2,4 = Transfer function regression constants; C4= 6,000; C1=1.00; and C2=1.00
*2
*1 2CC (Eq. 21)
856.2*2 1748.3940874.2 HMAHC (Eq. 22)
Where
HHMA = total HMA thickness (inches)
182
For the HMA alligator cracking model, the MEPDG allows the following model coefficients to be
modified as needed as part of Arizona local calibration to improve alligator cracking prediction
accuracy and remove bias:
C1,2,4 = Transfer function regression constants (C1=1.0; and C2=4.5, C4= 6,000;)
Transverse Cracking (Low Temperature Induced)
This model predicts transverse cracking in HMA pavement that is specifically caused by low‐
temperature contraction of asphalt binders that lead to tensile stresses and transverse crack
formation. Hourly HMA temperatures are calculated along with HMA properties such as creep
compliance to estimate tensile stress in the HMA surface. These cracks can be initially spaced at
more than 100 ft to as low as 20 ft (very severe) after a period of several years, and can
deteriorate and create significant roughness leading to increased IRI.
Transverse cracking prediction begins with characterizing the amount of crack propagation
induced by a given thermal cooling cycle using the Paris law of crack propagation (AASHTO
2008).
nC A K (Eq. 23)
Where
C = change in the crack depth due to a cooling cycle
K = change in the stress intensity factor due to a cooling cycle
A, n = fracture parameters for the HMA mixture
Experimental results indicate that reasonable estimates of A and n can be obtained from the
indirect tensile creep‐compliance and strength of the HMA in accordance with Eq. 24 and Eq. 25
(AASHTO 2008).
nELogk mHMAttA 52.2389.410 (Eq. 24)
And set
mn
118.0
183
Where
kt = coefficient determined through global calibration for each input level
(Level 1 = 5.0; Level 2 = 1.5; and Level 3 = 3.0)
EHMA = HMA indirect tensile modulus (psi)
m = mixture tensile strength (psi)
M = the m‐value derived from the indirect tensile creep compliance curve measured
in the laboratory
βt = local or mixture calibration factor
Stress intensity factor, K, was incorporated in the MEPDG through the use of a simplified
equation developed from theoretical finite element studies (Eq. 25).
56.099.145.0 otip CK (Eq. 25)
Where
tip = far‐field stress from pavement response model at depth of crack tip (psi)
Co = current crack length (ft)
The amount of transverse cracking is predicted by the MEPDG using an assumed relationship
between the probability distribution of the log of the crack depth to HMA layer thickness ratio
and the percent of cracking. Equation 26 shows the expression used to determine the amount of
thermal cracking (AASHTO 2008).
HMA
d
dt H
CLogNTC
1
1 (Eq. 26)
Where
TC = Thermal cracking (ft/mi)
βt1 = Regression coefficient determined through global calibration (400)
N[z] = Standard normal distribution evaluated at [z]
184
σd = Standard deviation of the log of the depth of cracks in the pavement
(0.769) (inches)
Cd = Crack depth (inches)
HHMA = Thickness of HMA layers (inches)
For the HMA transverse cracking model, the MEPDG allows the following model coefficients to
be modified as needed as part of local calibration to improve transverse cracking prediction
accuracy and remove bias:
βt1 = Regression coefficient determined through global calibration (400)
Rutting
This model predicts the amount of permanent deformation (rutting) in each pavement and
subgrade layer and accumulates total rutting. Rutting prediction in the MEPDG comprises
several submodels and algorithms for estimating incrementally critical vertical permanent or
plastic axial strain in the HMA and unbound aggregate bases materials and subgrade soils. The
MEPDG assumes rutting does not occur in chemically stabilized materials or layers.
Total rutting is the sum of all plastic vertical strain accumulated in each pavement layer due to
applied axle loading. Eq. 27 through Eq. 34 present the key models used by the MEPDG for
predicting rutting (AASHTO 2008):
HMA Rutting Model
For all HMA mixtures types, the MEPDG field calibrated form of the laboratory derived
relationship from repeated load permanent deformation tests is shown in Eq. 27.
rrrrr kkkHMArzrHMAHMApHMAp Tnkh 3322110)(1)()(
(Eq. 27)
Where
p(HMA) = accumulated permanent or plastic vertical deformation in the HMA layer or
sublayer (inches)
εp(HMA) = accumulated permanent or plastic axial strain in the HMA layer or sublayer
(inch/inch)
185
εr(HMA) = resilient or elastic strain calculated by the structural response model at the
mid‐depth of each HMA sublayer (inch/inch)
h(HMA) = thickness of the HMA layer/sublayer (inches)
n = number of axle load repetitions
T = mix or pavement temperature (° F)
kz = depth confinement factor
k1r,2r,3r = global field calibration parameters (from the NCHRP 1‐40D
recalibration; k1r = ‐3.35412, k2r =1.5606, k3r = 0.4791)
β1r, β2r, β3r,= local or mixture field calibration constants; for the global
calibration, these constants were all set to 1.0
Dz DCCk 328196.021 (Eq. 28)
342.174868.21039.0 21 HMAHMA HHC (Eq. 29)
428.277331.10172.0 22 HMAHMA HHC (Eq. 30)
Where
D = depth below the surface (inches)
HHMA = total HMA thickness (inches)
For the HMA rutting model, the MEPDG allows the following model coefficients to be modified
as part of the Arizona local calibration to improve rutting prediction accuracy and remove bias:
k1r,2r,3r = global field calibration parameters (from the NCHRP 1‐40D recalibration; k1r = ‐3.35412,
k2r =1.5606, k3r = 0.4791)
β1r, β2r, β3r, = Arizona local or mixture field calibration constants; for the global calibration, these
constants were all set to 0.69, 1.0, and 1.0.
186
Aggregate Base and Subgrade Soils Rutting Model
Eq. 31 shows the field‐calibrated mathematical equation used to calculate plastic vertical
deformation within all unbound pavement sublayers and the foundation or embankment soil.
n
r
osoilvsssoilp ehk 11)(
(Eq. 31)
Where
p(Soil) = permanent or plastic deformation for the layer or sublayer (inches)
n = number of axle load applications
o = intercept determined from laboratory repeated load permanent
deformation tests (inch/inch)
r = resilient strain imposed in laboratory test to obtain material properties εo, β,
and , in/in
v = average vertical resilient or elastic strain in the layer or sublayer, and
calculated by the structural response model (inch/inch)
hSoil = thickness of the unbound layer or sublayer (inches)
ks1 = global calibration coefficients; ks1=2.03 for granular materials and 1.35 for
fine‐grained materials
βs1 = local calibration constant for the rutting in the unbound layers (base or
subgrade); the local calibration constant was set to 1.0 for the global
calibration effort. βs1 represents the subgrade layer while βB1 represents the
base layer
cWLog 017638.061119.0
(Eq. 32)
1
9
9
10110
oC
(Eq. 33)
0075.09
1
9
1
br
br
oMa
MaLnC
(Eq. 34)
187
Where
Wc = water content (percent)
Mr = resilient modulus of the unbound layer or sublayer (psi)
a1,9 = regression constants; a1=0.15 and a9=20.0
b1,9 = regression constants; b1=0.0 and b9=0.0
For the unbound aggregate base and subgrade rutting model, the MEPDG allows the following
model coefficients to be modified as needed as part of Arizona local calibration to improve
rutting prediction accuracy and remove bias:
βs1 = local calibration constant for the rutting in the unbound layers (base or subgrade); the
Arizona local calibration constant was set to βs1 = 0.14 and βB1 = 0.37. βs1 represents the
subgrade layer while βB1 represents the base layer.
HMA Smoothness (IRI)
This model predicts the IRI in the traffic lane over time. HMA smoothness prediction in the
MEPDG is based on the principle that smoothness degradation (increasing IRI) from an initial as‐
constructed smoothness level is due mostly to the development of critical surface distresses and
subgrade foundation movement. Eq. 35 and Eq. 36 were developed using data from the LTPP
program to predict the IRI over time for new HMA pavements (AASHTO 2008).
RDCTCCFCCSFCIRIIRI Totalo 4321 (Eq. 35)
Where
IRIo = initial IRI after construction (inch/mi)
SF = site factor, refer to Eq. 23
FCTotal = area of fatigue cracking (combined alligator, longitudinal, and reflection
cracking in the wheel path) (percent of total lane area). All load‐related
cracks are combined on an area basis – length of cracks is multiplied by 1 ft
to convert length into an area basis
TC = length of transverse cracking (including the reflection of transverse cracks in
existing HMA pavements) (ft/mi).
RD = average rut depth (inches)
188
C1 = 0.0150
C2 = 0.4000
C3 = 0.0080
C4 = 40.000
SF = site factor
SF is calculated in accordance with the following equation:
SF = FROSTH + SWELLP*AGE1.5 (Eq. 36)
Where
FROSTH = LN([PRECIP+1]*FINES*[FI+1])
SWELLP = LN([PRECIP+1]*CLAY*[PI+1])
FINES = FSAND + SILT
AGE = pavement age (years)
PI = subgrade soil plasticity index
PRECIP = mean annual precipitation (inches)
FI = mean annual freezing index (° F days)
FSAND = amount of fine sand particles in subgrade (percent of particles between
0.074 and 0.42 mm)
SILT = amount of silt particles in subgrade (percent of particles between 0.074 and
0.002 mm)
CLAY = amount of clay size particles in subgrade (percent of particles less than
0.002 mm)
For the HMA IRI model, the MEPDG allows modifications to the following model coefficients as
part of the Arizona local calibration to improve IRI prediction accuracy and remove bias:
C1 = 1.2281
C2 = 0.1175
C3 = 0.0080
C4 = 0.0280
189
NEW AND RECONSTRUCTED JPCP
The JPCP MEPDG models developed originally under NCHRP Project 1‐37A have been
recalibrated twice since the completion of Project 1‐37A to improve global models prediction
accuracy and remove bias. Other reasons for recalibration included correcting systematic errors
in key MEPDG inputs that were discovered since the 2004 and 2007 completion of NCHRP
Projects 1‐37A and 1‐40D projects, respectively. The latest round of the JPCP models was
completed under NCHRP 20‐07 in 2011 using the correct CTE values. They are of interest in
Arizona since the correct CTE values are being used in the local calibration.
Transverse Slab Cracking
The MEPDG predicts both bottom‐up and top‐down modes of transverse slab cracking. The
estimates of bottom‐up and top‐down transverse slab cracking combined to predict total
transverse slab cracking. For both modes of cracking, the MEPDG estimates critical stresses at
the portland concrete cement (PCC) surface or bottom that is then related to an allowable
number of load repetitions before the slab cracks using empirical relationships. The allowable
number of load repetitions is then used for computing incrementally PCC top‐down and bottom‐
up fatigue damage by dividing the actual number of axle loads by the allowable number of axle
loads. Top‐down and bottom‐up transverse cracking is then estimated based on estimates of
fatigue damage.
The two estimates of slab cracking are combined in a manner that excludes the possibility of
both modes of cracking occurring on the same slab as under typical service conditions; the
potential for either mode of cracking is present in all slabs but although slabs may crack either
from the bottom up or top down, both do not occur simultaneously. Eq. 37 through Eq. 40
present the key models used by the MEPDG for predicting cracking (AASHTO 2008).
Top‐Down and Bottom‐Up Transverse Slab Cracking Model
541
1C
FDICCRK
(Eq. 37)
Where
CRK = predicted amount of bottom‐up or top‐down cracking (fraction).
DIF = fatigue damage calculated using the procedure described in this section.
C4 = 0.6 (NCHRP 20‐07)
C5 = ‐2.05 (NCHRP 20‐07)
190
For the JPCP cracking model, the MEPDG allows the following model coefficients to be modified
as needed as part of Arizona local calibration to improve cracking prediction accuracy and
remove bias:
C4 = 0.19
C5 = ‐2.067
PCC Fatigue Damage
The general expression for fatigue damage accumulations considering all critical factors for JPCP
transverse cracking follows (based on Miner’s hypothesis):
onmlkji
onmlkjiF N
nDI
,,,,,,
,,,,,, (Eq. 38)
Where
DIF = total fatigue damage (top‐down or bottom‐up)
ni,j,k, ... = applied number of load applications at condition i, j, k, l, m, n
Ni,j,k, …= allowable number of load applications at condition i, j, k, l, m, n
I = age (accounts for change in PCC modulus of rupture and elasticity, slab or base
contact friction, traffic loads)
j = month (accounts for change in base elastic modulus and effective dynamic
modulus of subgrade reaction)
k = axle type (single, tandem, and tridem for bottom‐up cracking; short,
medium, and long wheelbase for top‐down cracking)
l = load level (incremental load for each axle type)
m = equivalent temperature difference between top and bottom PCC surfaces
n = traffic offset path
o = hourly truck traffic fraction
191
Allowable Number of Load Repetitions
The applied number of load applications (ni,j,k,l,m,n) is the actual number of axle type k of load
level l that passed through traffic path n under each condition (age, season, and temperature
difference). The allowable number of load applications is the number of load cycles at which
fatigue failure is expected on average and is a function of the applied stress and PCC strength.
The allowable number of load applications is determined using the following globally calibrated
PCC fatigue equation:
2
,,,,,1,,,,,log
C
nmlkji
inmlkji
MRCN
(Eq. 39)
Where
Ni,j,k,… = allowable number of load applications at condition i, j, k, l, m, n
MRi = PCC modulus of rupture at age i (psi)
σi,j,k, = applied stress at condition i, j, k, l, m, n
C1 = calibration constant, 2.0
C2 = calibration constant, 1.22
For the JPCP transverse cracking allowable number of load repetitions model, the MEPDG allows
the following model coefficients to be modified as needed as part of Arizona local calibration to
improve cracking prediction accuracy and remove bias. However, these were not changed.
C1 = calibration constant, 2.0
C2 = calibration constant, 1.22
Total Transverse Cracking Model
The fatigue damage calculation is a process of summing damage from each damage increment.
Once top‐down and bottom‐up damage are estimated, the corresponding cracking is computed
using Eq. 37 and the total combined cracking determined using Eq. 40.
100 downTopupBottomdownTopupBottom CRKCRKCRKCRKTCRACK (Eq. 40)
192
Where
TCRACK = total transverse cracking (percent, all severities)
CRKBottop‐up = predicted amount of bottom‐up transverse cracking (fraction)
CRKTop‐down = predicted amount of top‐down transverse cracking (fraction)
Eq. 40 assumes that a slab may crack from either bottom up or top down, but not both.
Transverse Joint Faulting
The mean transverse joint faulting is predicted incrementally on a monthly basis. The magnitude
of increment is based on current faulting level, the number of axle loads applied, pavement
design features, material properties, and climatic conditions. Total faulting is determined as a
sum of faulting increments from all previous months (i.e., since traffic opening) using the
following equations (AASHTO 2008):
m
iim FaultFault
1
(Eq. 41)
iiii DEFaultFAULTMAXCFault *)(* 21134 (Eq. 42)
65
170 C *C CEROD
m
jji LogDEFAULTMAXFAULTMAX
(Eq. 43)
6
)*
(*)0.5*1(* *C 2005curling120
C
s
EROD
P
WetDaysPLogCLogFAULTMAX
(Eq. 44)
Where
Faultm = mean joint faulting at the end of month m (inches)
ΔFaulti = incremental change (monthly) in mean transverse joint faulting
during month i (inches)
FAULTMAXi = maximum mean transverse joint faulting for month I (inches)
FAULTMAX0 = initial maximum mean transverse joint faulting (inches)
EROD = base or subbase erodibility factor
193
DEi = differential density of energy of subgrade deformation accumulated during
month i
δcurling = maximum mean monthly slab corner upward deflection PCC due to
temperature curling and moisture warping
PS = overburden on subgrade (pounds)
P200 = percent subgrade material passing No. 200 sieve
WetDays = average annual number of wet days (greater than 0.1 inch rainfall)
C1 = 1.0184
C2 = 0.91656
C3 = 0.0021848
C4 = 0.000884
C5 = 250
C6 = 0.4
C7 = 1.83312
Note that C12 and C34 are defined by Eq. 45 and Eq. 46.
25.0
2112 *C CC FR (Eq. 45)
25.0
4334 *C CC FR (Eq. 46)
Where
FR = base freezing index defined as percentage of time the top base
temperature is below freezing (32° F) temperature.
194
For the JPCP transverse joint faulting model, the MEPDG allows the following model
coefficients to be modified as needed as part of the Arizona local calibration to improve
faulting prediction accuracy and remove bias:
C1 = 0.0355
C2 = 0.1147
C3 = 0.00436
C4 = 1.1*10‐07
C5 = 20000
C6 = 2.0389
C7 = 0.189
C8 = 400
Smoothness (IRI)
JPCP smoothness is predicted as a function of the initial as‐constructed smoothness and any
change in pavement longitudinal profile over time and traffic due to distress development and
progression and foundation movements. The IRI model was calibrated and validated using LTPP
data that represented variety of design, materials, foundations, and climatic conditions. The
following is the final globally calibrated model (AASHTO 2008):
IRI = IRII + C1*CRK +C2*SPALL + C3*TFAULT + C4*SF (Eq. 47)
Where
IRI = predicted IRI (inch/mi)
IRII = initial smoothness measured as IRI (inch/mi)
CRK = percent slabs with transverse cracks (all severities)
SPALL = percentage of joints with spalling (medium and high severities)
TFAULT = total joint faulting cumulated per mi (inches)
C1 = 0.8203
C2 = 0.4417
C3 = 0.4929
C4 = 25.24
SF = site factor
195
SF = AGE (1+0.5556*FI) (1+P200)*10‐6 (Eq. 48)
Where
AGE = pavement age (years)
FI = freezing index (° F days)
P200 = percent subgrade material passing No. 200 sieve
The transverse cracking and faulting are obtained using the MEPDG models described earlier.
The transverse joint spalling is determined in accordance with Eq. 49, which was calibrated using
LTPP and other data (AASHTO 2008):
SCF)AGE*(-12005.11
100
0.01AGE
AGESPALL (Eq. 49)
Where
SPALL = percentage joints spalled (medium and high severities)
AGE = pavement age since construction (years)
SCF = scaling factor based on site, design, and climate
SCF = –1400 + 350 • ACPCC • (0.5 + PREFORM) + 43.4 f'c ^ 0.4– 0.2 (FTcycles • AGE) +
43 HPCC – 536 WCPCC (Eq. 50)
Where
ACPCC = PCC air content (percent); 6 percent
AGE = time since construction (years)
PREFORM = 1 if preformed sealant is present; 0 if not
f'c = PCC compressive strength (psi)
FTcycles = average annual number of freeze‐thaw cycles
HPCC = PCC slab thickness (inches)
WCPCC = PCC water/cement ratio
196
For the JPCP IRI model, the MEPDG allows the following model coefficients to be modified as
part of the Arizona local calibration to improve IRI prediction accuracy and remove bias:
C1 = 0.60
C2 = 3.48
C3 = 1.22
C4 = 45.20
SUMMARY
Tables 66 and 67 present the local calibration coefficients for HMA pavements and JPCP,
respectively, that can be modified to improve model prediction accuracy and eliminate bias.
Table 66. HMA Pavement Local Calibration Coefficients
Distress/IRI Local Calibration Coefficients That
Improve Prediction Accuracy Local Calibration Coefficients That
Eliminate Bias
Rutting k1r,βS1,βB1 β2r, β3r, k 2r, k 3r
Alligator cracking C2, kf1 C1, kf2, kf3
Transverse cracking
βt1 βt1
IRI C2, C3, C4 C1
Table 67. JPCP Local Calibration Coefficients
Distress/IRI Local Calibration Coefficients That
Improve Prediction Accuracy Local Calibration Coefficients That
Eliminate Bias
Faulting C2, C3, C4, C5, C6, C7 C1
Transverse cracking
C2, C5 C1, C4
IRI C1, C2, C3 C4